diff --git a/Cargo.lock b/Cargo.lock
index b037fa2a7..608789855 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -1717,6 +1717,7 @@ dependencies = [
  "clap",
  "cli-table",
  "dialoguer",
+ "fabro-api",
  "fabro-macros",
  "fabro-model",
  "fabro-test",
@@ -1726,6 +1727,7 @@ dependencies = [
  "httpmock",
  "indicatif",
  "insta",
+ "progenitor-client",
  "rand 0.8.5",
  "reqwest 0.13.2",
  "serde",
diff --git a/docs/api-reference/fabro-api.yaml b/docs/api-reference/fabro-api.yaml
index 245082025..85d0a6c93 100644
--- a/docs/api-reference/fabro-api.yaml
+++ b/docs/api-reference/fabro-api.yaml
@@ -1443,6 +1443,8 @@ paths:
       summary: List Models
       description: Returns a paginated list of available LLM models from the built-in catalog.
       parameters:
+        - $ref: "#/components/parameters/ModelProviderFilter"
+        - $ref: "#/components/parameters/ModelQueryFilter"
         - $ref: "#/components/parameters/PageLimit"
         - $ref: "#/components/parameters/PageOffset"
       responses:
@@ -1452,6 +1454,12 @@ paths:
             application/json:
               schema:
                 $ref: "#/components/schemas/PaginatedModelList"
+        "400":
+          description: Invalid filter value
+          content:
+            application/json:
+              schema:
+                $ref: "#/components/schemas/ErrorResponse"
 
   /api/v1/models/{id}/test:
     post:
@@ -1466,6 +1474,7 @@ paths:
           schema:
             type: string
           description: The model identifier.
+        - $ref: "#/components/parameters/ModelTestModeParam"
       responses:
         "200":
           description: Test result
@@ -1473,6 +1482,12 @@ paths:
             application/json:
               schema:
                 $ref: "#/components/schemas/ModelTestResult"
+        "400":
+          description: Invalid test mode
+          content:
+            application/json:
+              schema:
+                $ref: "#/components/schemas/ErrorResponse"
         "404":
           description: Model not found
           content:
@@ -1751,6 +1766,33 @@ components:
         default: 0
       example: 0
 
+    ModelProviderFilter:
+      name: provider
+      in: query
+      required: false
+      description: Filter models by provider name. Invalid values return `400`.
+      schema:
+        type: string
+      example: anthropic
+
+    ModelQueryFilter:
+      name: query
+      in: query
+      required: false
+      description: Case-insensitive substring search across `id`, `display_name`, and `aliases`.
+      schema:
+        type: string
+      example: opus
+
+    ModelTestModeParam:
+      name: mode
+      in: query
+      required: false
+      description: Test mode for the single-model test endpoint. Defaults to `basic`.
+      schema:
+        $ref: "#/components/schemas/ModelTestMode"
+      example: basic
+
   schemas:
     # ── Pagination ───────────────────────────────────────────────────────
 
@@ -1951,7 +1993,7 @@ components:
           description: Whether this is the default model for its provider.
 
     ModelTestResult:
-      description: Result of testing a model with a simple prompt.
+      description: Result of testing a model in `basic` or `deep` mode.
       type: object
       required:
         - model_id
@@ -1972,6 +2014,13 @@ components:
           nullable: true
           description: Error details when status is "error".
 
+    ModelTestMode:
+      description: Single-model test mode.
+      type: string
+      enum:
+        - basic
+        - deep
+
     # ── Completion Schemas ─────────────────────────────────────────────
 
     CompletionMessage:
diff --git a/lib/crates/fabro-cli/src/args.rs b/lib/crates/fabro-cli/src/args.rs
index cf15026e1..b455668c0 100644
--- a/lib/crates/fabro-cli/src/args.rs
+++ b/lib/crates/fabro-cli/src/args.rs
@@ -4,7 +4,7 @@ use std::path::PathBuf;
 use clap::{Args, Subcommand, ValueEnum};
 use fabro_agent::cli::AgentArgs;
 use fabro_graphviz::render::GraphFormat;
-use fabro_llm::cli::{ChatArgs, ModelsCommand, PromptArgs};
+use fabro_llm::cli::ModelsCommand;
 
 pub(crate) const LONG_VERSION: &str = concat!(
     env!("CARGO_PKG_VERSION"),
@@ -731,9 +731,6 @@ impl RunsCommands {
 
 #[derive(Subcommand)]
 pub(crate) enum Commands {
-    /// LLM prompt operations
-    #[command(hide = true)]
-    Llm(LlmNamespace),
     /// Run an agentic coding session
     #[command(hide = true)]
     Exec(AgentArgs),
@@ -831,10 +828,6 @@ pub(crate) enum Commands {
 impl Commands {
     pub(crate) fn name(&self) -> &'static str {
         match self {
-            Self::Llm(ns) => match &ns.command {
-                LlmCommand::Prompt(_) => "llm prompt",
-                LlmCommand::Chat(_) => "llm chat",
-            },
             Self::Artifact(ns) => match &ns.command {
                 ArtifactCommand::List(_) => "artifact list",
                 ArtifactCommand::Cp(_) => "artifact cp",
@@ -1079,20 +1072,6 @@ pub(crate) struct CompletionArgs {
     pub shell: clap_complete::Shell,
 }
 
-#[derive(Args)]
-pub(crate) struct LlmNamespace {
-    #[command(subcommand)]
-    pub(crate) command: LlmCommand,
-}
-
-#[derive(Subcommand)]
-pub(crate) enum LlmCommand {
-    /// Execute a prompt
-    Prompt(PromptArgs),
-    /// Interactive multi-turn chat
-    Chat(ChatArgs),
-}
-
 #[derive(Args)]
 pub(crate) struct SkillNamespace {
     #[command(subcommand)]
diff --git a/lib/crates/fabro-cli/src/commands/llm/chat.rs b/lib/crates/fabro-cli/src/commands/llm/chat.rs
deleted file mode 100644
index b63587cec..000000000
--- a/lib/crates/fabro-cli/src/commands/llm/chat.rs
+++ /dev/null
@@ -1,39 +0,0 @@
-use anyhow::Result;
-use fabro_llm::cli::{ChatArgs, ServerConnection, run_chat, run_chat_via_server};
-use fabro_types::Settings;
-
-use crate::args::GlobalArgs;
-use crate::user_config::{ExecutionMode, build_server_client, resolve_mode};
-
-pub(super) async fn execute(
-    mut args: ChatArgs,
-    cli_settings: &Settings,
-    globals: &GlobalArgs,
-) -> Result<()> {
-    globals.require_no_json()?;
-    let llm_defaults = cli_settings.llm.as_ref();
-    if args.model.is_none() {
-        args.model = llm_defaults.and_then(|l| l.model.clone());
-    }
-
-    let resolved = resolve_mode(
-        globals.storage_dir.as_deref(),
-        globals.server_url.as_deref(),
-        cli_settings,
-    );
-    match resolved.mode {
-        ExecutionMode::Server => {
-            let client = build_server_client(resolved.tls.as_ref())?;
-            let server = ServerConnection {
-                client,
-                base_url: resolved.server_base_url,
-            };
-            run_chat_via_server(args, &server).await?;
-        }
-        ExecutionMode::Standalone => {
-            run_chat(args).await?;
-        }
-    }
-
-    Ok(())
-}
diff --git a/lib/crates/fabro-cli/src/commands/llm/mod.rs b/lib/crates/fabro-cli/src/commands/llm/mod.rs
deleted file mode 100644
index e86c603fe..000000000
--- a/lib/crates/fabro-cli/src/commands/llm/mod.rs
+++ /dev/null
@@ -1,16 +0,0 @@
-mod chat;
-mod prompt;
-
-use anyhow::Result;
-
-use crate::args::{GlobalArgs, LlmCommand, LlmNamespace};
-use crate::user_config::load_user_settings_with_globals;
-
-pub(crate) async fn dispatch(ns: LlmNamespace, globals: &GlobalArgs) -> Result<()> {
-    let cli_settings = load_user_settings_with_globals(globals)?;
-
-    match ns.command {
-        LlmCommand::Prompt(args) => prompt::execute(args, &cli_settings, globals).await,
-        LlmCommand::Chat(args) => chat::execute(args, &cli_settings, globals).await,
-    }
-}
diff --git a/lib/crates/fabro-cli/src/commands/llm/prompt.rs b/lib/crates/fabro-cli/src/commands/llm/prompt.rs
deleted file mode 100644
index e44c0a751..000000000
--- a/lib/crates/fabro-cli/src/commands/llm/prompt.rs
+++ /dev/null
@@ -1,38 +0,0 @@
-use anyhow::Result;
-use fabro_llm::cli::{PromptArgs, ServerConnection, run_prompt, run_prompt_via_server};
-use fabro_types::Settings;
-
-use crate::args::GlobalArgs;
-use crate::user_config::{ExecutionMode, build_server_client, resolve_mode};
-
-pub(super) async fn execute(
-    mut args: PromptArgs,
-    cli_settings: &Settings,
-    globals: &GlobalArgs,
-) -> Result<()> {
-    let llm_defaults = cli_settings.llm.as_ref();
-    if args.model.is_none() {
-        args.model = llm_defaults.and_then(|l| l.model.clone());
-    }
-
-    let resolved = resolve_mode(
-        globals.storage_dir.as_deref(),
-        globals.server_url.as_deref(),
-        cli_settings,
-    );
-    match resolved.mode {
-        ExecutionMode::Server => {
-            let client = build_server_client(resolved.tls.as_ref())?;
-            let server = ServerConnection {
-                client,
-                base_url: resolved.server_base_url,
-            };
-            run_prompt_via_server(args, &server, globals.json).await?;
-        }
-        ExecutionMode::Standalone => {
-            run_prompt(args, globals.json).await?;
-        }
-    }
-
-    Ok(())
-}
diff --git a/lib/crates/fabro-cli/src/commands/mod.rs b/lib/crates/fabro-cli/src/commands/mod.rs
index 2e4d171a6..aba2d0bfd 100644
--- a/lib/crates/fabro-cli/src/commands/mod.rs
+++ b/lib/crates/fabro-cli/src/commands/mod.rs
@@ -4,7 +4,6 @@ pub(crate) mod doctor;
 pub(crate) mod exec;
 pub(crate) mod graph;
 pub(crate) mod install;
-pub(crate) mod llm;
 pub(crate) mod model;
 pub(crate) mod parse;
 pub(crate) mod pr;
diff --git a/lib/crates/fabro-cli/src/commands/model.rs b/lib/crates/fabro-cli/src/commands/model.rs
index e4a42fbad..b46328d82 100644
--- a/lib/crates/fabro-cli/src/commands/model.rs
+++ b/lib/crates/fabro-cli/src/commands/model.rs
@@ -1,26 +1,22 @@
 use anyhow::Result;
-use fabro_llm::cli::{ModelsCommand, ServerConnection, run_models};
+use fabro_llm::cli::{ModelsCommand, run_models};
 
 use crate::args::GlobalArgs;
+use crate::server_client;
 use crate::user_config;
 
 pub(crate) async fn execute(command: Option<ModelsCommand>, globals: &GlobalArgs) -> Result<()> {
     let cli_settings = user_config::load_user_settings_with_globals(globals)?;
-    let resolved = user_config::resolve_mode(
-        globals.storage_dir.as_deref(),
-        globals.server_url.as_deref(),
-        &cli_settings,
-    );
-    let server = match resolved.mode {
-        user_config::ExecutionMode::Server => {
-            let client = user_config::build_server_client(resolved.tls.as_ref())?;
-            Some(ServerConnection {
-                client,
-                base_url: resolved.server_base_url,
-            })
+    let client = match globals.server_url.as_deref() {
+        Some(base_url) => {
+            let tls = cli_settings
+                .server
+                .as_ref()
+                .and_then(|server| server.tls.as_ref());
+            server_client::connect_remote_api_client(base_url, tls)?
         }
-        user_config::ExecutionMode::Standalone => None,
+        None => server_client::connect_api_client(&cli_settings.storage_dir()).await?,
     };
 
-    run_models(command, server, globals.json).await
+    run_models(command, client, globals.json).await
 }
diff --git a/lib/crates/fabro-cli/src/main.rs b/lib/crates/fabro-cli/src/main.rs
index 748f4146c..965231b80 100644
--- a/lib/crates/fabro-cli/src/main.rs
+++ b/lib/crates/fabro-cli/src/main.rs
@@ -160,7 +160,6 @@ async fn main_inner() -> (String, Result<()>) {
 
     let result = Box::pin(async move {
         match *command {
-            Commands::Llm(ns) => commands::llm::dispatch(ns, &globals).await?,
             Commands::Exec(args) => commands::exec::execute(args, &globals).await?,
             Commands::RunCmd(cmd) => Box::pin(commands::run::dispatch(cmd, &globals)).await?,
             Commands::Preflight(args) => commands::preflight::execute(args, &globals).await?,
diff --git a/lib/crates/fabro-cli/src/server_client.rs b/lib/crates/fabro-cli/src/server_client.rs
index 3fa4e39f9..628b24262 100644
--- a/lib/crates/fabro-cli/src/server_client.rs
+++ b/lib/crates/fabro-cli/src/server_client.rs
@@ -15,6 +15,7 @@ use serde::de::DeserializeOwned;
 use tokio::time::sleep;
 
 use crate::commands::server::start;
+use crate::user_config;
 
 pub(crate) struct ServerStoreClient {
     client: fabro_api::Client,
@@ -98,6 +99,12 @@ impl RunProjection {
 }
 
 pub(crate) async fn connect_server(storage_dir: &Path) -> Result<ServerStoreClient> {
+    Ok(ServerStoreClient {
+        client: connect_api_client(storage_dir).await?,
+    })
+}
+
+pub(crate) async fn connect_api_client(storage_dir: &Path) -> Result<fabro_api::Client> {
     let bind = start::ensure_server_running(storage_dir)
         .with_context(|| format!("Failed to start fabro server for {}", storage_dir.display()))?;
     let socket_path = match bind {
@@ -116,9 +123,18 @@ pub(crate) async fn connect_server(storage_dir: &Path) -> Result<ServerStoreClie
         .context("Failed to build Unix-socket HTTP client for fabro server")?;
     wait_for_server_ready(&http_client).await?;
 
-    Ok(ServerStoreClient {
-        client: fabro_api::Client::new_with_client("http://fabro", http_client),
-    })
+    Ok(fabro_api::Client::new_with_client(
+        "http://fabro",
+        http_client,
+    ))
+}
+
+pub(crate) fn connect_remote_api_client(
+    base_url: &str,
+    tls: Option<&user_config::ClientTlsSettings>,
+) -> Result<fabro_api::Client> {
+    let http_client = user_config::build_server_client(tls)?;
+    Ok(fabro_api::Client::new_with_client(base_url, http_client))
 }
 
 async fn wait_for_server_ready(http_client: &reqwest::Client) -> Result<()> {
diff --git a/lib/crates/fabro-cli/tests/it/cmd/fabro.rs b/lib/crates/fabro-cli/tests/it/cmd/fabro.rs
index 371aba9a6..3c5fa073f 100644
--- a/lib/crates/fabro-cli/tests/it/cmd/fabro.rs
+++ b/lib/crates/fabro-cli/tests/it/cmd/fabro.rs
@@ -59,3 +59,21 @@ fn help() {
     ----- stderr -----
     ");
 }
+
+#[test]
+fn llm_namespace_is_not_available() {
+    let context = test_context!();
+    let mut cmd = context.command();
+    cmd.arg("llm");
+    fabro_snapshot!(context.filters(), cmd, @"
+    success: false
+    exit_code: 2
+    ----- stdout -----
+    ----- stderr -----
+    error: unrecognized subcommand 'llm'
+
+    Usage: fabro [OPTIONS] <COMMAND>
+
+    For more information, try '--help'.
+    ");
+}
diff --git a/lib/crates/fabro-cli/tests/it/cmd/llm.rs b/lib/crates/fabro-cli/tests/it/cmd/llm.rs
deleted file mode 100644
index 1f8e7f2cc..000000000
--- a/lib/crates/fabro-cli/tests/it/cmd/llm.rs
+++ /dev/null
@@ -1,390 +0,0 @@
-use std::process::Output;
-
-use fabro_test::{TwinScenario, TwinScenarios, fabro_snapshot, test_context, twin_openai};
-use predicates::prelude::*;
-
-async fn run_success_output(mut cmd: assert_cmd::Command) -> Output {
-    tokio::task::spawn_blocking(move || cmd.assert().success().get_output().clone())
-        .await
-        .expect("blocking command task should complete")
-}
-
-#[test]
-fn prompt_bad_option() {
-    let context = test_context!();
-    let mut cmd = context.llm();
-    cmd.args(["prompt", "-o", "bad_option", "hello"]);
-    fabro_snapshot!(context.filters(), cmd, @"
-    success: false
-    exit_code: 2
-    ----- stdout -----
-    ----- stderr -----
-    error: invalid value 'bad_option' for '--option <OPTION>': expected key=value, got bad_option
-
-    For more information, try '--help'.
-    ");
-}
-
-#[test]
-fn prompt_no_text() {
-    let context = test_context!();
-    let mut cmd = context.llm();
-    cmd.arg("prompt");
-    fabro_snapshot!(context.filters(), cmd, @"
-    success: false
-    exit_code: 1
-    ----- stdout -----
-    ----- stderr -----
-    error: Error: no prompt provided. Pass a prompt as an argument or pipe text via stdin.
-    ");
-}
-
-#[test]
-fn prompt_schema_invalid() {
-    let context = test_context!();
-    let mut cmd = context.llm();
-    cmd.args([
-        "prompt",
-        "--no-stream",
-        "-m",
-        "test-model",
-        "--schema",
-        "not json",
-        "hello",
-    ]);
-    fabro_snapshot!(context.filters(), cmd, @"
-    success: false
-    exit_code: 1
-    ----- stdout -----
-    ----- stderr -----
-    Using model: test-model
-    error: --schema must be valid JSON
-      > expected ident at line 1 column 2
-    ");
-}
-
-#[test]
-fn prompt_reads_from_stdin() {
-    let context = test_context!();
-    let result = context
-        .llm()
-        .args(["prompt", "--no-stream", "-m", "test-model"])
-        .write_stdin("hello from stdin")
-        .assert()
-        .failure();
-
-    // Should NOT complain about missing prompt
-    result.stderr(predicate::str::contains("no prompt provided").not());
-}
-
-#[test]
-fn prompt_concatenates_stdin_and_arg() {
-    let context = test_context!();
-    let result = context
-        .llm()
-        .args([
-            "prompt",
-            "--no-stream",
-            "-m",
-            "test-model",
-            "summarize this",
-        ])
-        .write_stdin("some input text")
-        .assert()
-        .failure();
-
-    result.stderr(predicate::str::contains("no prompt provided").not());
-}
-
-#[fabro_macros::e2e_test(live("ANTHROPIC_API_KEY"))]
-fn prompt_no_stream_generates_response() {
-    let context = test_context!();
-    context
-        .llm()
-        .args([
-            "prompt",
-            "--no-stream",
-            "-m",
-            "claude-sonnet-4-5",
-            "Say just the word 'hello'",
-        ])
-        .assert()
-        .success()
-        .stdout(predicate::str::is_empty().not());
-}
-
-#[fabro_macros::e2e_test(twin)]
-async fn twin_prompt_no_stream() {
-    let context = test_context!();
-    let (base_url, api_key) = fabro_test::e2e_openai!();
-    let mut cmd = context.llm();
-    cmd.env("OPENAI_BASE_URL", base_url);
-    cmd.env("OPENAI_API_KEY", api_key);
-    cmd.args(["prompt", "--no-stream", "-m", "gpt-5.4-mini", "Say hello"]);
-    cmd.write_stdin("");
-    let output = run_success_output(cmd).await;
-    assert_eq!(
-        String::from_utf8(output.stdout).unwrap().trim(),
-        "deterministic: Say hello"
-    );
-}
-
-#[fabro_macros::e2e_test(live("ANTHROPIC_API_KEY"))]
-fn prompt_stream_generates_response() {
-    let context = test_context!();
-    context
-        .llm()
-        .args([
-            "prompt",
-            "-m",
-            "claude-sonnet-4-5",
-            "Say just the word 'hello'",
-        ])
-        .assert()
-        .success()
-        .stdout(predicate::str::is_empty().not());
-}
-
-#[fabro_macros::e2e_test(twin)]
-async fn twin_prompt_stream() {
-    let context = test_context!();
-    let (base_url, api_key) = fabro_test::e2e_openai!();
-    let mut cmd = context.llm();
-    cmd.env("OPENAI_BASE_URL", base_url);
-    cmd.env("OPENAI_API_KEY", api_key);
-    cmd.args(["prompt", "-m", "gpt-5.4-mini", "Say hello"]);
-    cmd.write_stdin("");
-    let output = run_success_output(cmd).await;
-    assert_eq!(
-        String::from_utf8(output.stdout).unwrap().trim(),
-        "deterministic: Say hello"
-    );
-}
-
-#[fabro_macros::e2e_test(live("ANTHROPIC_API_KEY"))]
-fn prompt_usage_shows_tokens() {
-    let context = test_context!();
-    context
-        .llm()
-        .args([
-            "prompt",
-            "--no-stream",
-            "-u",
-            "-m",
-            "claude-sonnet-4-5",
-            "Say just the word 'hello'",
-        ])
-        .assert()
-        .success()
-        .stderr(predicate::str::contains("Tokens:"));
-}
-
-#[fabro_macros::e2e_test(twin)]
-async fn twin_prompt_usage() {
-    let context = test_context!();
-    let (base_url, api_key) = fabro_test::e2e_openai!();
-    let mut cmd = context.llm();
-    cmd.env("OPENAI_BASE_URL", base_url);
-    cmd.env("OPENAI_API_KEY", api_key);
-    cmd.args([
-        "prompt",
-        "--no-stream",
-        "-u",
-        "-m",
-        "gpt-5.4-mini",
-        "Say hello",
-    ]);
-    cmd.write_stdin("");
-    let output = run_success_output(cmd).await;
-    assert_eq!(
-        String::from_utf8(output.stdout).unwrap().trim(),
-        "deterministic: Say hello"
-    );
-    assert!(
-        String::from_utf8(output.stderr)
-            .unwrap()
-            .contains("Tokens:"),
-        "stderr should include token usage"
-    );
-}
-
-#[fabro_macros::e2e_test(live("ANTHROPIC_API_KEY"))]
-fn prompt_schema_no_stream_generates_json() {
-    let context = test_context!();
-    let assert = context
-        .llm()
-        .args([
-            "prompt", "--no-stream", "-m", "claude-sonnet-4-5",
-            "--schema", r#"{"type":"object","properties":{"greeting":{"type":"string"}},"required":["greeting"]}"#,
-            "Return a JSON object with a greeting field set to hello",
-        ])
-        .assert()
-        .success();
-
-    let stdout = String::from_utf8(assert.get_output().stdout.clone()).unwrap();
-    let parsed: serde_json::Value =
-        serde_json::from_str(stdout.trim()).expect("stdout should be valid JSON");
-    assert!(
-        parsed.get("greeting").is_some(),
-        "expected 'greeting' key in output"
-    );
-}
-
-#[fabro_macros::e2e_test(twin)]
-async fn twin_prompt_schema_no_stream() {
-    let context = test_context!();
-    let twin = twin_openai().await;
-    let namespace = format!("{}::{}", module_path!(), line!());
-    TwinScenarios::new(namespace.clone())
-        .scenario(
-            TwinScenario::responses("gpt-5.4-mini")
-                .stream(false)
-                .input_contains("Return JSON")
-                .text(r#"{"greeting":"hello"}"#),
-        )
-        .load(twin)
-        .await;
-
-    let mut cmd = context.llm();
-    twin.configure_command(&mut cmd, &namespace);
-    cmd.args([
-        "prompt",
-        "--no-stream",
-        "-m",
-        "gpt-5.4-mini",
-        "--schema",
-        r#"{"type":"object","properties":{"greeting":{"type":"string"}},"required":["greeting"]}"#,
-        "Return JSON",
-    ]);
-    cmd.write_stdin("");
-    let output = run_success_output(cmd).await;
-
-    let stdout = String::from_utf8(output.stdout).unwrap();
-    let parsed: serde_json::Value =
-        serde_json::from_str(stdout.trim()).expect("stdout should be valid JSON");
-    assert_eq!(parsed["greeting"], "hello");
-}
-
-#[fabro_macros::e2e_test(live("ANTHROPIC_API_KEY"))]
-fn prompt_schema_stream_generates_json() {
-    let context = test_context!();
-    let assert = context
-        .llm()
-        .args([
-            "prompt", "-m", "claude-sonnet-4-5",
-            "--schema", r#"{"type":"object","properties":{"greeting":{"type":"string"}},"required":["greeting"]}"#,
-            "Return a JSON object with a greeting field set to hello",
-        ])
-        .assert()
-        .success();
-
-    let stdout = String::from_utf8(assert.get_output().stdout.clone()).unwrap();
-    let parsed: serde_json::Value =
-        serde_json::from_str(stdout.trim()).expect("stdout should be valid JSON");
-    assert!(
-        parsed.get("greeting").is_some(),
-        "expected 'greeting' key in output"
-    );
-}
-
-#[fabro_macros::e2e_test(twin)]
-async fn twin_prompt_schema_stream() {
-    let context = test_context!();
-    let twin = twin_openai().await;
-    let namespace = format!("{}::{}", module_path!(), line!());
-    TwinScenarios::new(namespace.clone())
-        .scenario(
-            TwinScenario::responses("gpt-5.4-mini")
-                .stream(true)
-                .input_contains("Return JSON")
-                .text(r#"{"greeting":"hello"}"#),
-        )
-        .load(twin)
-        .await;
-
-    let mut cmd = context.llm();
-    twin.configure_command(&mut cmd, &namespace);
-    cmd.args([
-        "prompt",
-        "-m",
-        "gpt-5.4-mini",
-        "--schema",
-        r#"{"type":"object","properties":{"greeting":{"type":"string"}},"required":["greeting"]}"#,
-        "Return JSON",
-    ]);
-    cmd.write_stdin("");
-    let output = run_success_output(cmd).await;
-
-    let stdout = String::from_utf8(output.stdout).unwrap();
-    let parsed: serde_json::Value =
-        serde_json::from_str(stdout.trim()).expect("stdout should be valid JSON");
-    assert_eq!(parsed["greeting"], "hello");
-}
-
-#[fabro_macros::e2e_test(live("ANTHROPIC_API_KEY"))]
-fn chat_multi_turn_with_system_prompt() {
-    let context = test_context!();
-    let assert = context
-        .command()
-        .args([
-            "llm",
-            "chat",
-            "-m",
-            "claude-haiku-4-5",
-            "-s",
-            "You are a pilot. End every response with 'Roger that.'",
-        ])
-        .write_stdin("What is your profession?\nWhat did I just ask you?\n")
-        .assert()
-        .success();
-
-    let stdout = String::from_utf8(assert.get_output().stdout.clone()).unwrap();
-    let stderr = String::from_utf8(assert.get_output().stderr.clone()).unwrap();
-
-    // Verify model info printed to stderr
-    assert!(
-        stderr.contains("Using model:"),
-        "stderr should show model info"
-    );
-
-    // Verify the system prompt influenced the output
-    assert!(
-        stdout.to_lowercase().contains("roger that"),
-        "response should follow pilot system prompt, got: {stdout}"
-    );
-
-    // Verify multi-turn: the second response should reference the first question
-    assert!(
-        stdout.to_lowercase().contains("profession")
-            || stdout.to_lowercase().contains("asked")
-            || stdout.to_lowercase().contains("pilot"),
-        "second response should show multi-turn context, got: {stdout}"
-    );
-}
-
-#[fabro_macros::e2e_test(twin)]
-async fn twin_chat_multi_turn() {
-    let context = test_context!();
-    let (base_url, api_key) = fabro_test::e2e_openai!();
-    let mut cmd = context.command();
-    cmd.env("OPENAI_BASE_URL", base_url);
-    cmd.env("OPENAI_API_KEY", api_key);
-    cmd.args([
-        "llm",
-        "chat",
-        "-m",
-        "gpt-5.4-mini",
-        "-s",
-        "You are a pilot. End every response with 'Roger that.'",
-    ]);
-    cmd.write_stdin("What is your profession?\nWhat did I just ask you?\n");
-    let output = run_success_output(cmd).await;
-
-    let stdout = String::from_utf8(output.stdout).unwrap();
-    let stderr = String::from_utf8(output.stderr).unwrap();
-    assert!(!stdout.trim().is_empty(), "stdout should not be empty");
-    assert!(
-        stderr.contains("Using model:"),
-        "stderr should show model info"
-    );
-}
diff --git a/lib/crates/fabro-cli/tests/it/cmd/llm_prompt.rs b/lib/crates/fabro-cli/tests/it/cmd/llm_prompt.rs
deleted file mode 100644
index b1f685be0..000000000
--- a/lib/crates/fabro-cli/tests/it/cmd/llm_prompt.rs
+++ /dev/null
@@ -1,87 +0,0 @@
-use fabro_test::{fabro_snapshot, test_context};
-use httpmock::prelude::*;
-use serde_json::Value;
-
-#[test]
-fn help() {
-    let context = test_context!();
-    let mut cmd = context.command();
-    cmd.args(["llm", "prompt", "--help"]);
-    fabro_snapshot!(context.filters(), cmd, @"
-    success: true
-    exit_code: 0
-    ----- stdout -----
-    Execute a prompt
-
-    Usage: fabro llm prompt [OPTIONS] [PROMPT]
-
-    Arguments:
-      [PROMPT]  The prompt text (also accepts stdin)
-
-    Options:
-          --json                       Output as JSON [env: FABRO_JSON=]
-      -m, --model <MODEL>              Model to use
-          --debug                      Enable DEBUG-level logging (default is INFO) [env: FABRO_DEBUG=]
-      -s, --system <SYSTEM>            System prompt
-          --no-stream                  Do not stream output
-          --no-upgrade-check           Disable automatic upgrade check [env: FABRO_NO_UPGRADE_CHECK=true]
-          --quiet                      Suppress non-essential output [env: FABRO_QUIET=]
-      -u, --usage                      Show token usage
-      -S, --schema <SCHEMA>            JSON schema for structured output (inline JSON string)
-          --verbose                    Enable verbose output [env: FABRO_VERBOSE=]
-      -o, --option <OPTION>            key=value options (temperature, `max_tokens`, `top_p`)
-          --storage-dir <STORAGE_DIR>  Storage directory (default: ~/.fabro) [env: FABRO_STORAGE_DIR=[STORAGE_DIR]]
-          --server-url <SERVER_URL>    Server URL (overrides server.base_url from user.toml) [env: FABRO_SERVER_URL=]
-      -h, --help                       Print help
-    ----- stderr -----
-    ");
-}
-
-#[test]
-fn prompt_json_streaming_server_reports_resolved_model() {
-    let context = test_context!();
-    let server = MockServer::start();
-    let sse_body = "\
-event: stream_event\n\
-data: {\"type\":\"text_delta\",\"delta\":\"Hi\",\"text_id\":null}\n\
-\n\
-event: stream_event\n\
-data: {\"type\":\"finish\",\"finish_reason\":\"stop\",\"usage\":{\"input_tokens\":5,\"output_tokens\":2,\"total_tokens\":7},\"response\":{\"id\":\"r1\",\"model\":\"resolved-model\",\"provider\":\"test-provider\",\"message\":{\"role\":\"assistant\",\"content\":[{\"kind\":\"text\",\"data\":\"Hi\"}],\"name\":null,\"tool_call_id\":null},\"finish_reason\":\"stop\",\"usage\":{\"input_tokens\":5,\"output_tokens\":2,\"total_tokens\":7},\"raw\":null,\"warnings\":[],\"rate_limit\":null}}\n\
-\n";
-
-    let mock = server.mock(|when, then| {
-        when.method(POST).path("/completions");
-        then.status(200)
-            .header("content-type", "text/event-stream")
-            .body(sse_body);
-    });
-
-    let output = context
-        .command()
-        .env_remove("FABRO_STORAGE_DIR")
-        .args([
-            "--server-url",
-            &server.url(""),
-            "--json",
-            "llm",
-            "prompt",
-            "Hello",
-        ])
-        .output()
-        .expect("command should run");
-
-    assert!(
-        output.status.success(),
-        "command failed:\nstdout:\n{}\nstderr:\n{}",
-        String::from_utf8_lossy(&output.stdout),
-        String::from_utf8_lossy(&output.stderr)
-    );
-
-    let value: Value =
-        serde_json::from_slice(&output.stdout).expect("llm prompt JSON should parse");
-    assert_eq!(value["response"], "Hi");
-    assert_eq!(value["model"], "resolved-model");
-    assert_eq!(value["usage"]["input_tokens"], 5);
-    assert_eq!(value["usage"]["output_tokens"], 2);
-    mock.assert();
-}
diff --git a/lib/crates/fabro-cli/tests/it/cmd/mod.rs b/lib/crates/fabro-cli/tests/it/cmd/mod.rs
index fa8683437..9d9c6c1a9 100644
--- a/lib/crates/fabro-cli/tests/it/cmd/mod.rs
+++ b/lib/crates/fabro-cli/tests/it/cmd/mod.rs
@@ -16,8 +16,6 @@ mod graph;
 mod inspect;
 mod install;
 mod json_global;
-mod llm;
-mod llm_prompt;
 mod logs;
 mod model;
 mod model_list;
diff --git a/lib/crates/fabro-cli/tests/it/cmd/model.rs b/lib/crates/fabro-cli/tests/it/cmd/model.rs
index 5899c19cf..a1e32ba66 100644
--- a/lib/crates/fabro-cli/tests/it/cmd/model.rs
+++ b/lib/crates/fabro-cli/tests/it/cmd/model.rs
@@ -161,3 +161,17 @@ fn list_query_case_insensitive() {
     ----- stderr -----
     ");
 }
+
+#[test]
+fn list_invalid_provider_errors() {
+    let context = test_context!();
+    let mut cmd = context.model();
+    cmd.args(["list", "--provider", "not-a-provider"]);
+    fabro_snapshot!(context.filters(), cmd, @"
+    success: false
+    exit_code: 1
+    ----- stdout -----
+    ----- stderr -----
+    error: unknown provider: not-a-provider
+    ");
+}
diff --git a/lib/crates/fabro-cli/tests/it/cmd/model_test.rs b/lib/crates/fabro-cli/tests/it/cmd/model_test.rs
index 743fdc340..9ae0987c5 100644
--- a/lib/crates/fabro-cli/tests/it/cmd/model_test.rs
+++ b/lib/crates/fabro-cli/tests/it/cmd/model_test.rs
@@ -40,6 +40,7 @@ fn model_test_unknown_model_errors() {
     exit_code: 1
     ----- stdout -----
     ----- stderr -----
+    Testing nonexistent-model-xyz... done
     error: Unknown model: nonexistent-model-xyz
     ");
 }
diff --git a/lib/crates/fabro-llm/Cargo.toml b/lib/crates/fabro-llm/Cargo.toml
index 636879597..aa19f37ac 100644
--- a/lib/crates/fabro-llm/Cargo.toml
+++ b/lib/crates/fabro-llm/Cargo.toml
@@ -36,6 +36,8 @@ indicatif.workspace = true
 clap.workspace = true
 dialoguer.workspace = true
 tracing.workspace = true
+progenitor-client = "0.13"
+fabro-api = { path = "../fabro-api" }
 fabro-model = { path = "../fabro-model" }
 fabro-util = { path = "../fabro-util" }
 
diff --git a/lib/crates/fabro-llm/src/cli.rs b/lib/crates/fabro-llm/src/cli.rs
index 44a5449f6..2465e91e4 100644
--- a/lib/crates/fabro-llm/src/cli.rs
+++ b/lib/crates/fabro-llm/src/cli.rs
@@ -2,24 +2,23 @@ use std::io::{self, IsTerminal, Read, Write};
 
 use dialoguer::console::Term;
 use dialoguer::theme::ColorfulTheme;
-use std::time::Duration;
 
-use anyhow::{Context, Result, bail};
+use anyhow::{Context, Result, anyhow, bail};
 use clap::{Args, Subcommand};
 use cli_table::format::{Border, Justify, Separator};
 use cli_table::{Cell, CellStruct, Color, Style, Table};
-use futures::{StreamExt, stream};
-use serde::{Deserialize, Serialize};
+use futures::StreamExt;
+use serde::Serialize;
+use serde::de::DeserializeOwned;
 use tokio::task;
-use tokio::time;
 
+use fabro_api::{self, types as api_types};
 use fabro_util::terminal::Styles;
 
 use fabro_model::{Catalog, Model, Provider};
 
 use crate::generate::{self, GenerateParams};
-use crate::tools::Tool;
-use crate::types::{ContentPart, GenerateResult, Message, ReasoningEffort, StreamEvent, Usage};
+use crate::types::{Message, StreamEvent, Usage};
 
 pub struct ServerConnection {
     pub client: reqwest::Client,
@@ -47,8 +46,6 @@ struct ModelTestRow {
 
 #[derive(Serialize)]
 struct ModelTestOutput {
-    #[serde(skip_serializing_if = "Option::is_none")]
-    deep_unsupported: Option<bool>,
     results: Vec<ModelTestRow>,
     total: usize,
     failures: u32,
@@ -932,186 +929,93 @@ pub async fn run_chat_via_server(args: ChatArgs, server: &ServerConnection) -> R
     Ok(())
 }
 
-#[derive(Deserialize)]
-struct PaginatedModelsResponse {
-    data: Vec<Model>,
+fn map_api_error<E>(err: progenitor_client::Error<E>) -> anyhow::Error
+where
+    E: serde::Serialize + std::fmt::Debug,
+{
+    match err {
+        progenitor_client::Error::ErrorResponse(response) => {
+            if let Ok(value) = serde_json::to_value(response.into_inner()) {
+                if let Some(detail) = value
+                    .get("errors")
+                    .and_then(serde_json::Value::as_array)
+                    .and_then(|errors| errors.first())
+                    .and_then(|entry| entry.get("detail"))
+                    .and_then(serde_json::Value::as_str)
+                {
+                    return anyhow!("{detail}");
+                }
+            }
+            anyhow!("request failed")
+        }
+        progenitor_client::Error::UnexpectedResponse(response) => {
+            anyhow!("request failed with status {}", response.status())
+        }
+        other => anyhow!("{other}"),
+    }
 }
 
-#[derive(Deserialize)]
-struct ModelTestResponse {
-    status: String,
-    error_message: Option<String>,
+fn convert_type<TInput, TOutput>(value: TInput) -> Result<TOutput>
+where
+    TInput: serde::Serialize,
+    TOutput: DeserializeOwned,
+{
+    serde_json::from_value(serde_json::to_value(value)?).map_err(Into::into)
 }
 
 async fn fetch_models_from_server(
-    client: &reqwest::Client,
-    base_url: &str,
+    client: &fabro_api::Client,
     provider: Option<&str>,
+    query: Option<&str>,
 ) -> Result<Vec<Model>> {
-    let url = format!("{base_url}/models?page[limit]=100");
-    tracing::debug!(url = %url, "Fetching models from server");
+    let mut offset = 0u64;
+    let mut models = Vec::new();
 
-    let response = client
-        .get(&url)
-        .send()
-        .await
-        .with_context(|| format!("Failed to connect to server at {base_url}"))?;
+    loop {
+        let mut request = client.list_models().page_limit(100u64).page_offset(offset);
+        if let Some(provider) = provider {
+            request = request.provider(provider.to_string());
+        }
+        if let Some(query) = query {
+            request = request.query(query.to_string());
+        }
 
-    let status = response.status();
-    if !status.is_success() {
-        let body = response.text().await.unwrap_or_default();
-        bail!("Server returned {status}: {body}");
-    }
-
-    let parsed: PaginatedModelsResponse = response
-        .json()
-        .await
-        .context("Failed to parse models response from server")?;
-
-    let mut models = parsed.data;
-    tracing::debug!(model_count = models.len(), "Models received from server");
-
-    if let Some(p) = provider {
-        models.retain(|m| m.provider.as_str() == p);
+        let response = request.send().await.map_err(map_api_error)?;
+        let parsed = response.into_inner();
+        let count = parsed.data.len() as u64;
+        models.extend(convert_type::<_, Vec<Model>>(parsed.data)?);
+        if !parsed.meta.has_more {
+            break;
+        }
+        offset += count;
     }
 
     Ok(models)
 }
 
 async fn test_model_via_server(
-    client: &reqwest::Client,
-    base_url: &str,
+    client: &fabro_api::Client,
     model_id: &str,
-) -> Result<ModelTestResponse> {
-    let url = format!("{base_url}/models/{model_id}/test");
-    let response = client
-        .post(&url)
-        .send()
-        .await
-        .with_context(|| format!("Failed to connect to server at {base_url}"))?;
-
-    let status = response.status();
-    if !status.is_success() {
-        let body = response.text().await.unwrap_or_default();
-        bail!("Server returned {status}: {body}");
+    mode: Option<api_types::ModelTestMode>,
+) -> Result<api_types::ModelTestResult> {
+    let mut request = client.test_model().id(model_id.to_string());
+    if let Some(mode) = mode {
+        request = request.mode(mode);
     }
-
-    response
-        .json()
-        .await
-        .context("Failed to parse model test response from server")
-}
-
-fn build_deep_test_params(info: &Model) -> Option<GenerateParams> {
-    if !info.features.tools {
-        return None;
-    }
-
-    let add_tool = Tool::active(
-        "add",
-        "Add two integers and return the sum",
-        serde_json::json!({
-            "type": "object",
-            "properties": {
-                "a": { "type": "integer", "description": "First number" },
-                "b": { "type": "integer", "description": "Second number" }
-            },
-            "required": ["a", "b"]
-        }),
-        |args, _ctx| async move {
-            let a = args
-                .get("a")
-                .and_then(serde_json::Value::as_i64)
-                .unwrap_or(0);
-            let b = args
-                .get("b")
-                .and_then(serde_json::Value::as_i64)
-                .unwrap_or(0);
-            Ok(serde_json::json!(a + b))
-        },
-    );
-
-    let mut params = GenerateParams::new(&info.id)
-        .provider(info.provider.as_str())
-        .prompt(
-            "I have three numbers: 15, 27, and 42. \
-             First use the add tool to compute 15 + 27, \
-             then use the add tool to add that result to 42. \
-             Finally, tell me whether the grand total is even or odd and why.",
-        )
-        .tools(vec![add_tool])
-        .max_tool_rounds(5)
-        .max_tokens(1024);
-
-    if info.features.reasoning {
-        params = params.reasoning_effort(ReasoningEffort::High);
-    }
-
-    Some(params)
-}
-
-fn validate_deep_result(result: &GenerateResult, info: &Model) -> (cli_table::Color, String) {
-    // Check tool use: need at least 2 steps (tool call + follow-up)
-    if result.steps.len() < 2 {
-        return (
-            Color::Red,
-            "deep: fail (model did not call tool)".to_string(),
-        );
-    }
-
-    // Check that step 0 had tool results (tool was executed)
-    if result.steps[0].tool_results.is_empty() {
-        return (Color::Red, "deep: fail (tool not executed)".to_string());
-    }
-
-    // Check correctness: 15+27=42, 42+42=84 — final response should contain "84"
-    let final_text = result.response.text();
-    if !final_text.contains("84") {
-        return (Color::Red, "deep: fail (wrong answer)".to_string());
-    }
-
-    // Check reasoning if the model supports it
-    if info.features.reasoning {
-        let has_reasoning = result.steps.iter().any(|step| {
-            step.response.message.content.iter().any(|part| {
-                matches!(part, ContentPart::Thinking(_))
-                    || matches!(part, ContentPart::Other { kind, .. } if kind == ContentPart::OPENAI_REASONING)
-            })
-        });
-        if !has_reasoning {
-            return (Color::Yellow, "deep: ok (no reasoning)".to_string());
-        }
-    }
-
-    (Color::Green, "deep: ok".to_string())
+    let response = request.send().await.map_err(map_api_error)?;
+    Ok(response.into_inner())
 }
 
 #[allow(clippy::print_stdout, clippy::print_stderr)]
 async fn test_models_via_server(
-    server: &ServerConnection,
+    client: &fabro_api::Client,
     provider: Option<&str>,
     model: Option<&str>,
     deep: bool,
     s: &Styles,
     json_output: bool,
 ) -> Result<()> {
-    if deep && !json_output {
-        eprintln!("Warning: --deep is not supported in server mode");
-    }
-    let models_to_test = if let Some(model_id) = model {
-        let all = fetch_models_from_server(&server.client, &server.base_url, None).await?;
-        let found: Vec<_> = all.into_iter().filter(|m| m.id == model_id).collect();
-        if found.is_empty() {
-            bail!("Unknown model: {model_id}");
-        }
-        found
-    } else {
-        fetch_models_from_server(&server.client, &server.base_url, provider).await?
-    };
-
-    if models_to_test.is_empty() {
-        bail!("No models found");
-    }
+    let request_mode = deep.then_some(api_types::ModelTestMode::Deep);
 
     let use_color = s.use_color;
     let mut title = models_title(use_color);
@@ -1120,31 +1024,47 @@ async fn test_models_via_server(
     let mut rows: Vec<Vec<CellStruct>> = Vec::new();
     let mut json_rows = Vec::new();
     let mut failures = 0u32;
-    for info in &models_to_test {
+    if let Some(model_id) = model {
         if !json_output {
-            eprint!("Testing {}...", info.id);
+            eprint!("Testing {model_id}...");
         }
-        let result = test_model_via_server(&server.client, &server.base_url, &info.id).await;
+        let result = test_model_via_server(client, model_id, request_mode).await;
         if !json_output {
             eprintln!(" done");
         }
 
-        let (result_color, status) = match result {
-            Ok(resp) if resp.status == "ok" => (Color::Green, "ok".to_string()),
+        let (info, result_color, status) = match result {
             Ok(resp) => {
-                failures += 1;
-                let msg = resp
-                    .error_message
-                    .unwrap_or_else(|| "unknown error".to_string());
-                (Color::Red, format!("error: {msg}"))
+                let info = Catalog::builtin()
+                    .get(&resp.model_id)
+                    .cloned()
+                    .with_context(|| {
+                        format!("Unknown model returned by server: {}", resp.model_id)
+                    })?;
+                if resp.status == api_types::ModelTestResultStatus::Ok {
+                    (info, Color::Green, "ok".to_string())
+                } else {
+                    failures += 1;
+                    let message = resp
+                        .error_message
+                        .unwrap_or_else(|| "unknown error".to_string());
+                    (info, Color::Red, format!("error: {message}"))
+                }
             }
-            Err(e) => {
+            Err(err) if err.to_string().contains("Model not found") => {
+                bail!("Unknown model: {model_id}");
+            }
+            Err(err) => {
+                let info = Catalog::builtin()
+                    .get(model_id)
+                    .cloned()
+                    .with_context(|| format!("Unknown model: {model_id}"))?;
                 failures += 1;
-                (Color::Red, format!("error: {e}"))
+                (info, Color::Red, format!("error: {err}"))
             }
         };
 
-        let mut row = model_row(info, use_color);
+        let mut row = model_row(&info, use_color);
         row.push(
             status
                 .clone()
@@ -1152,14 +1072,55 @@ async fn test_models_via_server(
                 .foreground_color(color_if(use_color, result_color)),
         );
         rows.push(row);
-        json_rows.push(model_test_row_from_status(info, &status, result_color));
+        json_rows.push(model_test_row_from_status(&info, &status, result_color));
+    } else {
+        let models_to_test = fetch_models_from_server(client, provider, None).await?;
+        if models_to_test.is_empty() {
+            bail!("No models found");
+        }
+
+        for info in &models_to_test {
+            if !json_output {
+                eprint!("Testing {}...", info.id);
+            }
+            let result = test_model_via_server(client, &info.id, request_mode).await;
+            if !json_output {
+                eprintln!(" done");
+            }
+
+            let (result_color, status) = match result {
+                Ok(resp) if resp.status == api_types::ModelTestResultStatus::Ok => {
+                    (Color::Green, "ok".to_string())
+                }
+                Ok(resp) => {
+                    failures += 1;
+                    let message = resp
+                        .error_message
+                        .unwrap_or_else(|| "unknown error".to_string());
+                    (Color::Red, format!("error: {message}"))
+                }
+                Err(err) => {
+                    failures += 1;
+                    (Color::Red, format!("error: {err}"))
+                }
+            };
+
+            let mut row = model_row(info, use_color);
+            row.push(
+                status
+                    .clone()
+                    .cell()
+                    .foreground_color(color_if(use_color, result_color)),
+            );
+            rows.push(row);
+            json_rows.push(model_test_row_from_status(info, &status, result_color));
+        }
     }
 
     if json_output {
         println!(
             "{}",
             serde_json::to_string_pretty(&ModelTestOutput {
-                deep_unsupported: deep.then_some(true),
                 total: json_rows.len(),
                 failures,
                 results: json_rows,
@@ -1189,7 +1150,7 @@ async fn test_models_via_server(
 #[allow(clippy::print_stdout)]
 pub async fn run_models(
     command: Option<ModelsCommand>,
-    server: Option<ServerConnection>,
+    client: fabro_api::Client,
     json_output: bool,
 ) -> Result<()> {
     let command = command.unwrap_or(ModelsCommand::List {
@@ -1201,23 +1162,8 @@ pub async fn run_models(
 
     match command {
         ModelsCommand::List { provider, query } => {
-            let mut models = if let Some(s) = &server {
-                fetch_models_from_server(&s.client, &s.base_url, provider.as_deref()).await?
-            } else {
-                let p = provider.as_deref().and_then(|s| s.parse::<Provider>().ok());
-                Catalog::builtin().list(p).into_iter().cloned().collect()
-            };
-
-            if let Some(q) = &query {
-                let q_lower = q.to_lowercase();
-                models.retain(|m| {
-                    m.id.to_lowercase().contains(&q_lower)
-                        || m.display_name.to_lowercase().contains(&q_lower)
-                        || m.aliases
-                            .iter()
-                            .any(|a| a.to_lowercase().contains(&q_lower))
-                });
-            }
+            let models =
+                fetch_models_from_server(&client, provider.as_deref(), query.as_deref()).await?;
 
             if json_output {
                 println!("{}", serde_json::to_string_pretty(&models)?);
@@ -1229,181 +1175,17 @@ pub async fn run_models(
             provider,
             model,
             deep,
-        } => match &server {
-            Some(s) => {
-                test_models_via_server(
-                    s,
-                    provider.as_deref(),
-                    model.as_deref(),
-                    deep,
-                    &styles,
-                    json_output,
-                )
-                .await?;
-            }
-            None => {
-                test_models(
-                    provider.as_deref(),
-                    model.as_deref(),
-                    deep,
-                    &styles,
-                    json_output,
-                )
-                .await?;
-            }
-        },
-    }
-
-    Ok(())
-}
-
-async fn test_one_model(info: &Model, deep: bool) -> (Color, String) {
-    if deep {
-        match build_deep_test_params(info) {
-            None => (Color::Yellow, "deep: skipped (no tool support)".to_string()),
-            Some(params) => {
-                let result =
-                    time::timeout(Duration::from_secs(90), generate::generate(params)).await;
-                match result {
-                    Ok(Ok(ref gen_result)) => validate_deep_result(gen_result, info),
-                    Ok(Err(e)) => (Color::Red, format!("deep: error: {e}")),
-                    Err(_) => (Color::Red, "deep: error: timeout (90s)".to_string()),
-                }
-            }
+        } => {
+            test_models_via_server(
+                &client,
+                provider.as_deref(),
+                model.as_deref(),
+                deep,
+                &styles,
+                json_output,
+            )
+            .await?;
         }
-    } else {
-        let params = GenerateParams::new(&info.id)
-            .provider(info.provider.as_str())
-            .prompt("Say OK")
-            .max_tokens(16);
-
-        let result = time::timeout(Duration::from_secs(30), generate::generate(params)).await;
-        match result {
-            Ok(Ok(_)) => (Color::Green, "ok".to_string()),
-            Ok(Err(e)) => (Color::Red, format!("error: {e}")),
-            Err(_) => (Color::Red, "error: timeout (30s)".to_string()),
-        }
-    }
-}
-
-#[allow(clippy::print_stdout)]
-async fn test_models(
-    provider: Option<&str>,
-    model: Option<&str>,
-    deep: bool,
-    s: &Styles,
-    json_output: bool,
-) -> Result<()> {
-    use rand::seq::SliceRandom;
-
-    let models_to_test = if let Some(model_id) = model {
-        match Catalog::builtin().get(model_id) {
-            Some(info) => vec![info.clone()],
-            None => bail!("Unknown model: {model_id}"),
-        }
-    } else {
-        let p = provider.and_then(|s| s.parse::<Provider>().ok());
-        Catalog::builtin().list(p).into_iter().cloned().collect()
-    };
-
-    if models_to_test.is_empty() {
-        bail!("No models found");
-    }
-
-    let pb = if json_output {
-        None
-    } else {
-        let test_kind = if deep { "Deep testing" } else { "Testing" };
-        let pb = indicatif::ProgressBar::new(models_to_test.len() as u64);
-        pb.set_style(
-            indicatif::ProgressStyle::with_template(&format!(
-                "{{spinner:.green}} {test_kind} {{pos}}/{{len}} models {{wide_bar}} {{eta}}"
-            ))
-            .unwrap(),
-        );
-        pb.enable_steady_tick(Duration::from_millis(100));
-        Some(pb)
-    };
-
-    // Build (original_index, model_info) pairs, then shuffle for provider spread
-    let mut indexed: Vec<(usize, &Model)> = models_to_test.iter().enumerate().collect();
-    indexed.shuffle(&mut rand::thread_rng());
-
-    // Run tests concurrently, 6 at a time
-    let results: Vec<(usize, Color, String)> = stream::iter(indexed)
-        .map(|(idx, info)| {
-            let pb = pb.clone();
-            async move {
-                let (color, status) = test_one_model(info, deep).await;
-                if let Some(pb) = pb {
-                    pb.inc(1);
-                }
-                (idx, color, status)
-            }
-        })
-        .buffer_unordered(6)
-        .collect()
-        .await;
-
-    if let Some(pb) = pb {
-        pb.finish_and_clear();
-    }
-
-    // Sort back to original catalog order
-    let mut sorted_results = results;
-    sorted_results.sort_by_key(|(idx, _, _)| *idx);
-
-    let use_color = s.use_color;
-    let mut title = models_title(use_color);
-    title.push("RESULT".cell().bold(use_color));
-
-    let mut rows: Vec<Vec<CellStruct>> = Vec::new();
-    let mut json_rows = Vec::new();
-    let mut failures = 0u32;
-    for (idx, result_color, status) in &sorted_results {
-        if *result_color == Color::Red {
-            failures += 1;
-        }
-        json_rows.push(model_test_row_from_status(
-            &models_to_test[*idx],
-            status,
-            *result_color,
-        ));
-        let mut row = model_row(&models_to_test[*idx], use_color);
-        row.push(
-            status
-                .cell()
-                .foreground_color(color_if(use_color, *result_color)),
-        );
-        rows.push(row);
-    }
-
-    if json_output {
-        println!(
-            "{}",
-            serde_json::to_string_pretty(&ModelTestOutput {
-                deep_unsupported: None,
-                total: json_rows.len(),
-                failures,
-                results: json_rows,
-            })?
-        );
-        if failures > 0 {
-            bail!("{failures} model(s) failed");
-        }
-        return Ok(());
-    }
-
-    let table = rows
-        .table()
-        .title(title)
-        .color_choice(color_choice(use_color))
-        .border(Border::builder().build())
-        .separator(Separator::builder().build());
-    println!("{}", table.display()?);
-
-    if failures > 0 {
-        bail!("{failures} model(s) failed");
     }
 
     Ok(())
@@ -1412,11 +1194,46 @@ async fn test_models(
 #[cfg(test)]
 mod tests {
     use super::*;
+    use fabro_model::{ModelCosts, ModelFeatures, ModelLimits, Provider};
 
     fn test_http_client() -> reqwest::Client {
         reqwest::Client::builder().no_proxy().build().unwrap()
     }
 
+    fn test_api_client(base_url: &str) -> fabro_api::Client {
+        fabro_api::Client::new_with_client(base_url, test_http_client())
+    }
+
+    fn test_model_json(id: &str, provider: Provider) -> serde_json::Value {
+        serde_json::to_value(Model {
+            id: id.to_string(),
+            provider,
+            family: "test".to_string(),
+            display_name: format!("{id} display"),
+            limits: ModelLimits {
+                context_window: 128_000,
+                max_output: Some(4096),
+            },
+            training: None,
+            knowledge_cutoff: None,
+            features: ModelFeatures {
+                tools: true,
+                vision: false,
+                reasoning: false,
+                effort: false,
+            },
+            costs: ModelCosts {
+                input_cost_per_mtok: Some(1.0),
+                output_cost_per_mtok: Some(2.0),
+                cache_input_cost_per_mtok: None,
+            },
+            estimated_output_tps: Some(100.0),
+            aliases: vec!["tm".to_string()],
+            default: false,
+        })
+        .unwrap()
+    }
+
     // --- parse_option ---
 
     #[test]
@@ -1609,7 +1426,7 @@ mod tests {
         let server = httpmock::MockServer::start_async().await;
         server
             .mock_async(|when, then| {
-                when.method("POST").path("/models/test-model/test");
+                when.method("POST").path("/api/v1/models/test-model/test");
                 then.status(200)
                     .header("Content-Type", "application/json")
                     .body(
@@ -1622,21 +1439,23 @@ mod tests {
             })
             .await;
 
-        let client = test_http_client();
-        let resp = test_model_via_server(&client, &server.url(""), "test-model")
+        let client = test_api_client(&server.url(""));
+        let resp = test_model_via_server(&client, "test-model", None)
             .await
             .unwrap();
 
-        assert_eq!(resp.status, "ok");
+        assert_eq!(resp.status, api_types::ModelTestResultStatus::Ok);
         assert!(resp.error_message.is_none());
     }
 
     #[tokio::test]
-    async fn test_model_via_server_parses_error() {
+    async fn test_model_via_server_passes_mode_and_parses_error() {
         let server = httpmock::MockServer::start_async().await;
         server
             .mock_async(|when, then| {
-                when.method("POST").path("/models/test-model/test");
+                when.method("POST")
+                    .path("/api/v1/models/test-model/test")
+                    .query_param("mode", "deep");
                 then.status(200)
                     .header("Content-Type", "application/json")
                     .body(
@@ -1650,12 +1469,13 @@ mod tests {
             })
             .await;
 
-        let client = test_http_client();
-        let resp = test_model_via_server(&client, &server.url(""), "test-model")
-            .await
-            .unwrap();
+        let client = test_api_client(&server.url(""));
+        let resp =
+            test_model_via_server(&client, "test-model", Some(api_types::ModelTestMode::Deep))
+                .await
+                .unwrap();
 
-        assert_eq!(resp.status, "error");
+        assert_eq!(resp.status, api_types::ModelTestResultStatus::Error);
         assert_eq!(resp.error_message.as_deref(), Some("timeout"));
     }
 
@@ -1664,7 +1484,7 @@ mod tests {
         let server = httpmock::MockServer::start_async().await;
         server
             .mock_async(|when, then| {
-                when.method("POST").path("/models/bad-model/test");
+                when.method("POST").path("/api/v1/models/bad-model/test");
                 then.status(404)
                 .header("Content-Type", "application/json")
                 .body(serde_json::json!({
@@ -1673,9 +1493,10 @@ mod tests {
             })
             .await;
 
-        let client = test_http_client();
-        let result = test_model_via_server(&client, &server.url(""), "bad-model").await;
+        let client = test_api_client(&server.url(""));
+        let result = test_model_via_server(&client, "bad-model", None).await;
         assert!(result.is_err());
+        assert!(result.unwrap_err().to_string().contains("Model not found"));
     }
 
     // --- fetch_models_from_server ---
@@ -1683,32 +1504,26 @@ mod tests {
     #[tokio::test]
     async fn fetch_models_from_server_parses_response() {
         let server = httpmock::MockServer::start_async().await;
-        let mock = server.mock_async(|when, then| {
-            when.method("GET").path("/models").query_param("page[limit]", "100");
-            then.status(200)
-                .header("Content-Type", "application/json")
-                .body(serde_json::json!({
-                    "data": [{
-                        "id": "test-model",
-                        "provider": "anthropic",
-                        "family": "test",
-                        "display_name": "Test Model",
-                        "limits": { "context_window": 128_000, "max_output": 4096 },
-                        "training": null,
-                        "features": { "tools": true, "vision": false, "reasoning": false },
-                        "costs": { "input_cost_per_mtok": 1.0, "output_cost_per_mtok": 2.0, "cache_input_cost_per_mtok": null },
-                        "estimated_output_tps": 100.0,
-                        "aliases": ["tm"],
-                        "default": false
-                    }],
-                    "meta": { "has_more": false }
-                }).to_string());
-        }).await;
+        let mock = server
+            .mock_async(|when, then| {
+                when.method("GET")
+                    .path("/api/v1/models")
+                    .query_param("page[limit]", "100")
+                    .query_param("page[offset]", "0");
+                then.status(200)
+                    .header("Content-Type", "application/json")
+                    .body(
+                        serde_json::json!({
+                            "data": [test_model_json("test-model", Provider::Anthropic)],
+                            "meta": { "has_more": false }
+                        })
+                        .to_string(),
+                    );
+            })
+            .await;
 
-        let client = test_http_client();
-        let models = fetch_models_from_server(&client, &server.url(""), None)
-            .await
-            .unwrap();
+        let client = test_api_client(&server.url(""));
+        let models = fetch_models_from_server(&client, None, None).await.unwrap();
 
         mock.assert_async().await;
         assert_eq!(models.len(), 1);
@@ -1721,44 +1536,25 @@ mod tests {
         let server = httpmock::MockServer::start_async().await;
         server
             .mock_async(|when, then| {
-                when.method("GET").path("/models");
+                when.method("GET")
+                    .path("/api/v1/models")
+                    .query_param("page[limit]", "100")
+                    .query_param("page[offset]", "0")
+                    .query_param("provider", "anthropic");
                 then.status(200)
                     .header("Content-Type", "application/json")
                     .body(
-                    serde_json::json!({
-                        "data": [
-                            {
-                                "id": "model-a",
-                                "provider": "anthropic",
-                                "family": "a",
-                                "display_name": "Model A",
-                                "limits": { "context_window": 8000 },
-                                "features": { "tools": false, "vision": false, "reasoning": false },
-                                "costs": {},
-                                "aliases": [],
-                                "default": false
-                            },
-                            {
-                                "id": "model-b",
-                                "provider": "openai",
-                                "family": "b",
-                                "display_name": "Model B",
-                                "limits": { "context_window": 8000 },
-                                "features": { "tools": false, "vision": false, "reasoning": false },
-                                "costs": {},
-                                "aliases": [],
-                                "default": false
-                            }
-                        ],
-                        "meta": { "has_more": false }
-                    })
-                    .to_string(),
-                );
+                        serde_json::json!({
+                            "data": [test_model_json("model-a", Provider::Anthropic)],
+                            "meta": { "has_more": false }
+                        })
+                        .to_string(),
+                    );
             })
             .await;
 
-        let client = test_http_client();
-        let models = fetch_models_from_server(&client, &server.url(""), Some("anthropic"))
+        let client = test_api_client(&server.url(""));
+        let models = fetch_models_from_server(&client, Some("anthropic"), None)
             .await
             .unwrap();
 
@@ -1766,18 +1562,101 @@ mod tests {
         assert_eq!(models[0].id, "model-a");
     }
 
+    #[tokio::test]
+    async fn fetch_models_from_server_passes_query_param() {
+        let server = httpmock::MockServer::start_async().await;
+        let mock = server
+            .mock_async(|when, then| {
+                when.method("GET")
+                    .path("/api/v1/models")
+                    .query_param("page[limit]", "100")
+                    .query_param("page[offset]", "0")
+                    .query_param("query", "sonnet");
+                then.status(200)
+                    .header("Content-Type", "application/json")
+                    .body(
+                        serde_json::json!({
+                            "data": [test_model_json("claude-sonnet-4-5", Provider::Anthropic)],
+                            "meta": { "has_more": false }
+                        })
+                        .to_string(),
+                    );
+            })
+            .await;
+
+        let client = test_api_client(&server.url(""));
+        let models = fetch_models_from_server(&client, None, Some("sonnet"))
+            .await
+            .unwrap();
+
+        mock.assert_async().await;
+        assert_eq!(models.len(), 1);
+        assert_eq!(models[0].id, "claude-sonnet-4-5");
+    }
+
+    #[tokio::test]
+    async fn fetch_models_from_server_follows_pagination() {
+        let server = httpmock::MockServer::start_async().await;
+        let first_page = server
+            .mock_async(|when, then| {
+                when.method("GET")
+                    .path("/api/v1/models")
+                    .query_param("page[limit]", "100")
+                    .query_param("page[offset]", "0");
+                then.status(200)
+                    .header("Content-Type", "application/json")
+                    .body(
+                        serde_json::json!({
+                            "data": [test_model_json("model-a", Provider::Anthropic)],
+                            "meta": { "has_more": true }
+                        })
+                        .to_string(),
+                    );
+            })
+            .await;
+        let second_page = server
+            .mock_async(|when, then| {
+                when.method("GET")
+                    .path("/api/v1/models")
+                    .query_param("page[limit]", "100")
+                    .query_param("page[offset]", "1");
+                then.status(200)
+                    .header("Content-Type", "application/json")
+                    .body(
+                        serde_json::json!({
+                            "data": [test_model_json("model-b", Provider::OpenAi)],
+                            "meta": { "has_more": false }
+                        })
+                        .to_string(),
+                    );
+            })
+            .await;
+
+        let client = test_api_client(&server.url(""));
+        let models = fetch_models_from_server(&client, None, None).await.unwrap();
+
+        first_page.assert_async().await;
+        second_page.assert_async().await;
+        assert_eq!(models.len(), 2);
+        assert_eq!(models[0].id, "model-a");
+        assert_eq!(models[1].id, "model-b");
+    }
+
     #[tokio::test]
     async fn fetch_models_from_server_error_on_failure() {
         let server = httpmock::MockServer::start_async().await;
         server
             .mock_async(|when, then| {
-                when.method("GET").path("/models");
+                when.method("GET")
+                    .path("/api/v1/models")
+                    .query_param("page[limit]", "100")
+                    .query_param("page[offset]", "0");
                 then.status(500).body("internal error");
             })
             .await;
 
-        let client = test_http_client();
-        let result = fetch_models_from_server(&client, &server.url(""), None).await;
+        let client = test_api_client(&server.url(""));
+        let result = fetch_models_from_server(&client, None, None).await;
         assert!(result.is_err());
     }
 
diff --git a/lib/crates/fabro-llm/src/lib.rs b/lib/crates/fabro-llm/src/lib.rs
index bf028fdce..91f10b6ae 100644
--- a/lib/crates/fabro-llm/src/lib.rs
+++ b/lib/crates/fabro-llm/src/lib.rs
@@ -3,6 +3,7 @@ pub mod client;
 pub mod error;
 pub mod generate;
 pub mod middleware;
+pub mod model_test;
 pub mod provider;
 pub mod providers;
 pub mod retry;
diff --git a/lib/crates/fabro-llm/src/model_test.rs b/lib/crates/fabro-llm/src/model_test.rs
new file mode 100644
index 000000000..390f54e91
--- /dev/null
+++ b/lib/crates/fabro-llm/src/model_test.rs
@@ -0,0 +1,284 @@
+use std::str::FromStr;
+use std::time::Duration;
+
+use tokio::time;
+
+use crate::generate::{self, GenerateParams};
+use crate::tools::Tool;
+use crate::types::GenerateResult;
+use crate::types::ReasoningEffort;
+use fabro_model::Model;
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Default)]
+pub enum ModelTestMode {
+    #[default]
+    Basic,
+    Deep,
+}
+
+impl ModelTestMode {
+    #[must_use]
+    pub const fn as_str(self) -> &'static str {
+        match self {
+            Self::Basic => "basic",
+            Self::Deep => "deep",
+        }
+    }
+
+    #[must_use]
+    pub const fn timeout_secs(self) -> u64 {
+        match self {
+            Self::Basic => 30,
+            Self::Deep => 90,
+        }
+    }
+}
+
+impl FromStr for ModelTestMode {
+    type Err = String;
+
+    fn from_str(value: &str) -> Result<Self, Self::Err> {
+        match value {
+            "basic" => Ok(Self::Basic),
+            "deep" => Ok(Self::Deep),
+            other => Err(format!("invalid model test mode: {other}")),
+        }
+    }
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum ModelTestStatus {
+    Ok,
+    Error,
+}
+
+impl ModelTestStatus {
+    #[must_use]
+    pub const fn as_str(self) -> &'static str {
+        match self {
+            Self::Ok => "ok",
+            Self::Error => "error",
+        }
+    }
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct ModelTestOutcome {
+    pub status: ModelTestStatus,
+    pub error_message: Option<String>,
+}
+
+impl ModelTestOutcome {
+    #[must_use]
+    pub fn ok() -> Self {
+        Self {
+            status: ModelTestStatus::Ok,
+            error_message: None,
+        }
+    }
+
+    #[must_use]
+    pub fn error(message: impl Into<String>) -> Self {
+        Self {
+            status: ModelTestStatus::Error,
+            error_message: Some(message.into()),
+        }
+    }
+}
+
+pub async fn run_model_test(info: &Model, mode: ModelTestMode) -> ModelTestOutcome {
+    match mode {
+        ModelTestMode::Basic => run_basic_test(info).await,
+        ModelTestMode::Deep => run_deep_test(info).await,
+    }
+}
+
+async fn run_basic_test(info: &Model) -> ModelTestOutcome {
+    let params = GenerateParams::new(&info.id)
+        .provider(info.provider.as_str())
+        .prompt("Say OK")
+        .max_tokens(16);
+
+    let result = time::timeout(
+        Duration::from_secs(ModelTestMode::Basic.timeout_secs()),
+        generate::generate(params),
+    )
+    .await;
+
+    match result {
+        Ok(Ok(_)) => ModelTestOutcome::ok(),
+        Ok(Err(err)) => ModelTestOutcome::error(err.to_string()),
+        Err(_) => ModelTestOutcome::error("timeout (30s)"),
+    }
+}
+
+async fn run_deep_test(info: &Model) -> ModelTestOutcome {
+    let Some(params) = build_deep_test_params(info) else {
+        return ModelTestOutcome::error("model does not support tools");
+    };
+
+    let result = time::timeout(
+        Duration::from_secs(ModelTestMode::Deep.timeout_secs()),
+        generate::generate(params),
+    )
+    .await;
+
+    match result {
+        Ok(Ok(gen_result)) => match validate_deep_result(&gen_result) {
+            Ok(()) => ModelTestOutcome::ok(),
+            Err(message) => ModelTestOutcome::error(message),
+        },
+        Ok(Err(err)) => ModelTestOutcome::error(err.to_string()),
+        Err(_) => ModelTestOutcome::error("timeout (90s)"),
+    }
+}
+
+fn build_deep_test_params(info: &Model) -> Option<GenerateParams> {
+    if !info.features.tools {
+        return None;
+    }
+
+    let add_tool = Tool::active(
+        "add",
+        "Add two integers and return the sum",
+        serde_json::json!({
+            "type": "object",
+            "properties": {
+                "a": { "type": "integer", "description": "First number" },
+                "b": { "type": "integer", "description": "Second number" }
+            },
+            "required": ["a", "b"]
+        }),
+        |args, _ctx| async move {
+            let a = args
+                .get("a")
+                .and_then(serde_json::Value::as_i64)
+                .unwrap_or(0);
+            let b = args
+                .get("b")
+                .and_then(serde_json::Value::as_i64)
+                .unwrap_or(0);
+            Ok(serde_json::json!(a + b))
+        },
+    );
+
+    let mut params = GenerateParams::new(&info.id)
+        .provider(info.provider.as_str())
+        .prompt(
+            "Use the add tool twice: first add 15 and 27, then add that result to 42. \
+             Finally, tell me whether the grand total is even or odd and why.",
+        )
+        .tools(vec![add_tool])
+        .max_tool_rounds(5)
+        .max_tokens(1024);
+
+    if info.features.reasoning {
+        params = params.reasoning_effort(ReasoningEffort::High);
+    }
+
+    Some(params)
+}
+
+fn validate_deep_result(result: &GenerateResult) -> Result<(), String> {
+    if result.steps.len() < 2 {
+        return Err("model did not call tool".to_string());
+    }
+
+    if result.steps[0].tool_results.is_empty() {
+        return Err("tool was not executed".to_string());
+    }
+
+    if !result.response.text().contains("84") {
+        return Err("wrong answer".to_string());
+    }
+
+    Ok(())
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::types::{FinishReason, Message, Response, StepResult, Usage};
+    use fabro_model::{ModelCosts, ModelFeatures, ModelLimits, Provider};
+
+    fn test_model_with(features: ModelFeatures) -> Model {
+        Model {
+            id: "test-model".to_string(),
+            provider: Provider::Anthropic,
+            family: "test".to_string(),
+            display_name: "Test Model".to_string(),
+            limits: ModelLimits {
+                context_window: 200_000,
+                max_output: Some(8_000),
+            },
+            training: None,
+            knowledge_cutoff: None,
+            features,
+            costs: ModelCosts {
+                input_cost_per_mtok: None,
+                output_cost_per_mtok: None,
+                cache_input_cost_per_mtok: None,
+            },
+            estimated_output_tps: None,
+            aliases: vec![],
+            default: false,
+        }
+    }
+
+    fn response_with_text(text: &str) -> Response {
+        Response {
+            id: "resp_1".to_string(),
+            model: "test-model".to_string(),
+            provider: "anthropic".to_string(),
+            message: Message::assistant(text),
+            finish_reason: FinishReason::Stop,
+            usage: Usage::default(),
+            raw: None,
+            warnings: vec![],
+            rate_limit: None,
+        }
+    }
+
+    #[tokio::test]
+    async fn run_model_test_deep_errors_when_model_lacks_tools() {
+        let info = test_model_with(ModelFeatures {
+            tools: false,
+            vision: false,
+            reasoning: true,
+            effort: true,
+        });
+
+        let outcome = run_model_test(&info, ModelTestMode::Deep).await;
+
+        assert_eq!(outcome.status, ModelTestStatus::Error);
+        assert_eq!(
+            outcome.error_message.as_deref(),
+            Some("model does not support tools")
+        );
+    }
+
+    #[test]
+    fn validate_deep_result_does_not_fail_only_for_missing_reasoning() {
+        let tool_results = vec![crate::types::ToolResult::success(
+            "call_1",
+            serde_json::json!(42),
+        )];
+        let first_step = StepResult {
+            response: response_with_text("tool step"),
+            tool_results: tool_results.clone(),
+        };
+        let second_step = StepResult {
+            response: response_with_text("84 is even"),
+            tool_results: vec![],
+        };
+        let result = GenerateResult {
+            response: response_with_text("84 is even"),
+            tool_results,
+            total_usage: Usage::default(),
+            steps: vec![first_step, second_step],
+            output: None,
+        };
+
+        assert_eq!(validate_deep_result(&result), Ok(()));
+    }
+}
diff --git a/lib/crates/fabro-server/src/demo/mod.rs b/lib/crates/fabro-server/src/demo/mod.rs
index c26e53f79..ab382549e 100644
--- a/lib/crates/fabro-server/src/demo/mod.rs
+++ b/lib/crates/fabro-server/src/demo/mod.rs
@@ -578,23 +578,6 @@ pub(crate) async fn list_query_history(
     paginated_response(insights::history(), &pagination)
 }
 
-// ── Models ────────────────────────────────────────────────────────────
-
-pub(crate) async fn list_models(
-    _auth: AuthenticatedService,
-    State(_state): State<Arc<AppState>>,
-    Query(pagination): Query<PaginationParams>,
-) -> Response {
-    paginated_response(
-        fabro_model::Catalog::builtin()
-            .list(None)
-            .into_iter()
-            .cloned()
-            .collect::<Vec<_>>(),
-        &pagination,
-    )
-}
-
 // ── Settings ───────────────────────────────────────────────────────────
 
 pub(crate) async fn get_server_settings(
diff --git a/lib/crates/fabro-server/src/server.rs b/lib/crates/fabro-server/src/server.rs
index 97af17332..7607f91b5 100644
--- a/lib/crates/fabro-server/src/server.rs
+++ b/lib/crates/fabro-server/src/server.rs
@@ -16,7 +16,8 @@ use axum::{Json, Router};
 use axum_extra::extract::cookie::Key;
 use bytes::Bytes;
 use fabro_llm::client::Client as LlmClient;
-use fabro_llm::generate::{GenerateParams, generate, generate_object};
+use fabro_llm::generate::{GenerateParams, generate_object};
+use fabro_llm::model_test::{ModelTestMode, run_model_test};
 use fabro_llm::types::{
     ContentPart, FinishReason, Message as LlmMessage, Request as LlmRequest,
     Response as LlmResponse, Role, StreamEvent, ToolChoice, ToolDefinition, Usage,
@@ -32,7 +33,7 @@ use tokio::sync::broadcast;
 use tokio::sync::oneshot;
 use tokio::sync::{Notify, OnceCell};
 use tokio::task::spawn_blocking;
-use tokio::time::{sleep, timeout};
+use tokio::time::sleep;
 use tokio_stream::StreamExt;
 use tower::{ServiceExt, service_fn};
 use ulid::Ulid;
@@ -77,6 +78,24 @@ pub struct PaginationParams {
     pub offset: u32,
 }
 
+#[derive(serde::Deserialize)]
+struct ModelListParams {
+    #[serde(rename = "page[limit]", default = "default_page_limit")]
+    limit: u32,
+    #[serde(rename = "page[offset]", default)]
+    offset: u32,
+    #[serde(default)]
+    provider: Option<String>,
+    #[serde(default)]
+    query: Option<String>,
+}
+
+#[derive(serde::Deserialize)]
+struct ModelTestParams {
+    #[serde(default)]
+    mode: Option<String>,
+}
+
 #[derive(serde::Deserialize)]
 struct EventListParams {
     #[serde(default)]
@@ -325,7 +344,7 @@ fn demo_routes() -> Router<Arc<AppState>> {
         )
         .route("/insights/execute", post(demo::execute_query_stub))
         .route("/insights/history", get(demo::list_query_history))
-        .route("/models", get(demo::list_models))
+        .route("/models", get(list_models))
         .route("/models/{id}/test", post(test_model))
         .route("/completions", post(create_completion))
         .route("/settings", get(demo::get_server_settings))
@@ -405,7 +424,7 @@ fn real_routes() -> Router<Arc<AppState>> {
         )
         .route("/insights/execute", post(not_implemented))
         .route("/insights/history", get(not_implemented))
-        .route("/models", get(demo::list_models))
+        .route("/models", get(list_models))
         .route("/models/{id}/test", post(test_model))
         .route("/completions", post(create_completion))
         .route("/settings", get(not_implemented))
@@ -1818,49 +1837,85 @@ async fn unpause_run(
     }
 }
 
+async fn list_models(
+    _auth: AuthenticatedService,
+    State(_state): State<Arc<AppState>>,
+    Query(params): Query<ModelListParams>,
+) -> Response {
+    let provider = match params.provider.as_deref() {
+        Some(value) => match fabro_model::Provider::from_str(value) {
+            Ok(provider) => Some(provider),
+            Err(err) => return ApiError::new(StatusCode::BAD_REQUEST, err).into_response(),
+        },
+        None => None,
+    };
+
+    let query = params.query.as_ref().map(|value| value.to_lowercase());
+    let limit = params.limit.clamp(1, 100) as usize;
+    let offset = params.offset as usize;
+
+    let mut models = fabro_model::Catalog::builtin()
+        .list(provider)
+        .into_iter()
+        .filter(|model| match &query {
+            Some(query) => {
+                model.id.to_lowercase().contains(query)
+                    || model.display_name.to_lowercase().contains(query)
+                    || model
+                        .aliases
+                        .iter()
+                        .any(|alias| alias.to_lowercase().contains(query))
+            }
+            None => true,
+        })
+        .cloned()
+        .collect::<Vec<_>>();
+
+    let has_more = models.len() > offset.saturating_add(limit);
+    let data = models.drain(offset..models.len().min(offset.saturating_add(limit)));
+
+    (
+        StatusCode::OK,
+        Json(serde_json::json!({
+            "data": data.collect::<Vec<_>>(),
+            "meta": { "has_more": has_more }
+        })),
+    )
+        .into_response()
+}
+
 async fn test_model(
     _auth: AuthenticatedService,
     State(state): State<Arc<AppState>>,
     Path(id): Path<String>,
+    Query(params): Query<ModelTestParams>,
 ) -> Response {
+    let mode = match params.mode.as_deref() {
+        Some(value) => match ModelTestMode::from_str(value) {
+            Ok(mode) => mode,
+            Err(err) => return ApiError::new(StatusCode::BAD_REQUEST, err).into_response(),
+        },
+        None => ModelTestMode::Basic,
+    };
     let Some(info) = fabro_model::Catalog::builtin().get(&id) else {
         return ApiError::not_found(format!("Model not found: {id}")).into_response();
     };
 
     if state.dry_run() {
         return Json(serde_json::json!({
-            "model_id": id,
+            "model_id": info.id,
             "status": "ok",
         }))
         .into_response();
     }
 
-    let params = GenerateParams::new(&info.id)
-        .provider(info.provider.as_str())
-        .prompt("Say OK")
-        .max_tokens(16);
-
-    let result = timeout(Duration::from_secs(30), generate(params)).await;
-
-    match result {
-        Ok(Ok(_)) => Json(serde_json::json!({
-            "model_id": id,
-            "status": "ok",
-        }))
-        .into_response(),
-        Ok(Err(e)) => Json(serde_json::json!({
-            "model_id": id,
-            "status": "error",
-            "error_message": e.to_string(),
-        }))
-        .into_response(),
-        Err(_) => Json(serde_json::json!({
-            "model_id": id,
-            "status": "error",
-            "error_message": "timeout (30s)",
-        }))
-        .into_response(),
-    }
+    let outcome = run_model_test(info, mode).await;
+    Json(serde_json::json!({
+        "model_id": info.id,
+        "status": outcome.status.as_str(),
+        "error_message": outcome.error_message,
+    }))
+    .into_response()
 }
 
 fn finish_reason_to_api_stop_reason(reason: &FinishReason) -> String {
@@ -2346,6 +2401,109 @@ mod tests {
         assert!(body["status"] == "ok" || body["status"] == "error");
     }
 
+    #[tokio::test]
+    async fn test_model_alias_returns_canonical_model_id() {
+        let state = create_app_state_with_options(dry_run_settings(), 5);
+        let app = build_router(state, AuthMode::Disabled);
+
+        let req = Request::builder()
+            .method("POST")
+            .uri(api("/models/sonnet/test"))
+            .header("content-type", "application/json")
+            .body(Body::empty())
+            .unwrap();
+
+        let response = app.oneshot(req).await.unwrap();
+        assert_eq!(response.status(), StatusCode::OK);
+
+        let body = body_json(response.into_body()).await;
+        assert_eq!(body["model_id"], "claude-sonnet-4-6");
+        assert_eq!(body["status"], "ok");
+    }
+
+    #[tokio::test]
+    async fn test_model_invalid_mode_returns_400() {
+        let state = create_app_state_with_options(dry_run_settings(), 5);
+        let app = build_router(state, AuthMode::Disabled);
+
+        let req = Request::builder()
+            .method("POST")
+            .uri(api("/models/claude-opus-4-6/test?mode=bogus"))
+            .header("content-type", "application/json")
+            .body(Body::empty())
+            .unwrap();
+
+        let response = app.oneshot(req).await.unwrap();
+        assert_eq!(response.status(), StatusCode::BAD_REQUEST);
+    }
+
+    #[tokio::test]
+    async fn list_models_filters_by_provider() {
+        let app = test_app_with();
+
+        let req = Request::builder()
+            .method("GET")
+            .uri(api("/models?provider=anthropic"))
+            .body(Body::empty())
+            .unwrap();
+
+        let response = app.oneshot(req).await.unwrap();
+        assert_eq!(response.status(), StatusCode::OK);
+
+        let body = body_json(response.into_body()).await;
+        let models = body["data"].as_array().unwrap();
+        assert!(!models.is_empty());
+        assert!(
+            models
+                .iter()
+                .all(|model| model["provider"] == serde_json::Value::String("anthropic".into()))
+        );
+    }
+
+    #[tokio::test]
+    async fn list_models_filters_by_query_across_aliases() {
+        let app = test_app_with();
+
+        let req = Request::builder()
+            .method("GET")
+            .uri(api("/models?query=codex"))
+            .body(Body::empty())
+            .unwrap();
+
+        let response = app.oneshot(req).await.unwrap();
+        assert_eq!(response.status(), StatusCode::OK);
+
+        let body = body_json(response.into_body()).await;
+        let model_ids = body["data"]
+            .as_array()
+            .unwrap()
+            .iter()
+            .map(|model| model["id"].as_str().unwrap().to_string())
+            .collect::<Vec<_>>();
+        assert_eq!(
+            model_ids,
+            vec![
+                "gpt-5.2-codex".to_string(),
+                "gpt-5.3-codex".to_string(),
+                "gpt-5.3-codex-spark".to_string()
+            ]
+        );
+    }
+
+    #[tokio::test]
+    async fn list_models_invalid_provider_returns_400() {
+        let app = test_app_with();
+
+        let req = Request::builder()
+            .method("GET")
+            .uri(api("/models?provider=not-a-provider"))
+            .body(Body::empty())
+            .unwrap();
+
+        let response = app.oneshot(req).await.unwrap();
+        assert_eq!(response.status(), StatusCode::BAD_REQUEST);
+    }
+
     #[tokio::test]
     async fn auth_login_github_redirects_to_github() {
         let mut settings = Settings::default();
diff --git a/lib/crates/fabro-test/src/lib.rs b/lib/crates/fabro-test/src/lib.rs
index 1f0793f33..c051cf1cf 100644
--- a/lib/crates/fabro-test/src/lib.rs
+++ b/lib/crates/fabro-test/src/lib.rs
@@ -561,13 +561,6 @@ impl TestContext {
         cmd
     }
 
-    /// Build a `llm` subcommand.
-    pub fn llm(&self) -> Command {
-        let mut cmd = self.command();
-        cmd.arg("llm");
-        cmd
-    }
-
     /// Build a `settings` subcommand.
     pub fn settings(&self) -> Command {
         let mut cmd = self.command();
diff --git a/lib/packages/fabro-api-client/src/.openapi-generator/FILES b/lib/packages/fabro-api-client/src/.openapi-generator/FILES
index 42ca3193c..92ea826d4 100644
--- a/lib/packages/fabro-api-client/src/.openapi-generator/FILES
+++ b/lib/packages/fabro-api-client/src/.openapi-generator/FILES
@@ -86,6 +86,7 @@ models/model-costs.ts
 models/model-features.ts
 models/model-limits.ts
 models/model-reference.ts
+models/model-test-mode.ts
 models/model-test-result.ts
 models/model.ts
 models/node-state.ts
diff --git a/lib/packages/fabro-api-client/src/api/models-api.ts b/lib/packages/fabro-api-client/src/api/models-api.ts
index e21cf1ae5..e53066ecf 100644
--- a/lib/packages/fabro-api-client/src/api/models-api.ts
+++ b/lib/packages/fabro-api-client/src/api/models-api.ts
@@ -24,6 +24,8 @@ import { BASE_PATH, COLLECTION_FORMATS, type RequestArgs, BaseAPI, RequiredError
 // @ts-ignore
 import type { ErrorResponse } from '../models';
 // @ts-ignore
+import type { ModelTestMode } from '../models';
+// @ts-ignore
 import type { ModelTestResult } from '../models';
 // @ts-ignore
 import type { PaginatedModelList } from '../models';
@@ -35,12 +37,14 @@ export const ModelsApiAxiosParamCreator = function (configuration?: Configuratio
         /**
          * Returns a paginated list of available LLM models from the built-in catalog.
          * @summary List Models
+         * @param {string} [provider] Filter models by provider name. Invalid values return &#x60;400&#x60;.
+         * @param {string} [query] Case-insensitive substring search across &#x60;id&#x60;, &#x60;display_name&#x60;, and &#x60;aliases&#x60;.
          * @param {number} [pageLimit] Maximum number of items to return per page.
          * @param {number} [pageOffset] Number of items to skip before returning results.
          * @param {*} [options] Override http request option.
          * @throws {RequiredError}
          */
-        listModels: async (pageLimit?: number, pageOffset?: number, options: RawAxiosRequestConfig = {}): Promise<RequestArgs> => {
+        listModels: async (provider?: string, query?: string, pageLimit?: number, pageOffset?: number, options: RawAxiosRequestConfig = {}): Promise<RequestArgs> => {
             const localVarPath = `/api/v1/models`;
             // use dummy base URL string because the URL constructor only accepts absolute URLs.
             const localVarUrlObj = new URL(localVarPath, DUMMY_BASE_URL);
@@ -60,6 +64,14 @@ export const ModelsApiAxiosParamCreator = function (configuration?: Configuratio
             // http bearer authentication required
             await setBearerAuthToObject(localVarHeaderParameter, configuration)
 
+            if (provider !== undefined) {
+                localVarQueryParameter['provider'] = provider;
+            }
+
+            if (query !== undefined) {
+                localVarQueryParameter['query'] = query;
+            }
+
             if (pageLimit !== undefined) {
                 localVarQueryParameter['page[limit]'] = pageLimit;
             }
@@ -83,10 +95,11 @@ export const ModelsApiAxiosParamCreator = function (configuration?: Configuratio
          * Tests a model by sending a simple prompt and reporting pass/fail.
          * @summary Test Model
          * @param {string} id The model identifier.
+         * @param {ModelTestMode} [mode] Test mode for the single-model test endpoint. Defaults to &#x60;basic&#x60;.
          * @param {*} [options] Override http request option.
          * @throws {RequiredError}
          */
-        testModel: async (id: string, options: RawAxiosRequestConfig = {}): Promise<RequestArgs> => {
+        testModel: async (id: string, mode?: ModelTestMode, options: RawAxiosRequestConfig = {}): Promise<RequestArgs> => {
             // verify required parameter 'id' is not null or undefined
             assertParamExists('testModel', 'id', id)
             const localVarPath = `/api/v1/models/{id}/test`
@@ -109,6 +122,10 @@ export const ModelsApiAxiosParamCreator = function (configuration?: Configuratio
             // http bearer authentication required
             await setBearerAuthToObject(localVarHeaderParameter, configuration)
 
+            if (mode !== undefined) {
+                localVarQueryParameter['mode'] = mode;
+            }
+
             localVarHeaderParameter['Accept'] = 'application/json';
 
             setSearchParams(localVarUrlObj, localVarQueryParameter);
@@ -132,13 +149,15 @@ export const ModelsApiFp = function(configuration?: Configuration) {
         /**
          * Returns a paginated list of available LLM models from the built-in catalog.
          * @summary List Models
+         * @param {string} [provider] Filter models by provider name. Invalid values return &#x60;400&#x60;.
+         * @param {string} [query] Case-insensitive substring search across &#x60;id&#x60;, &#x60;display_name&#x60;, and &#x60;aliases&#x60;.
          * @param {number} [pageLimit] Maximum number of items to return per page.
          * @param {number} [pageOffset] Number of items to skip before returning results.
          * @param {*} [options] Override http request option.
          * @throws {RequiredError}
          */
-        async listModels(pageLimit?: number, pageOffset?: number, options?: RawAxiosRequestConfig): Promise<(axios?: AxiosInstance, basePath?: string) => AxiosPromise<PaginatedModelList>> {
-            const localVarAxiosArgs = await localVarAxiosParamCreator.listModels(pageLimit, pageOffset, options);
+        async listModels(provider?: string, query?: string, pageLimit?: number, pageOffset?: number, options?: RawAxiosRequestConfig): Promise<(axios?: AxiosInstance, basePath?: string) => AxiosPromise<PaginatedModelList>> {
+            const localVarAxiosArgs = await localVarAxiosParamCreator.listModels(provider, query, pageLimit, pageOffset, options);
             const localVarOperationServerIndex = configuration?.serverIndex ?? 0;
             const localVarOperationServerBasePath = operationServerMap['ModelsApi.listModels']?.[localVarOperationServerIndex]?.url;
             return (axios, basePath) => createRequestFunction(localVarAxiosArgs, globalAxios, BASE_PATH, configuration)(axios, localVarOperationServerBasePath || basePath);
@@ -147,11 +166,12 @@ export const ModelsApiFp = function(configuration?: Configuration) {
          * Tests a model by sending a simple prompt and reporting pass/fail.
          * @summary Test Model
          * @param {string} id The model identifier.
+         * @param {ModelTestMode} [mode] Test mode for the single-model test endpoint. Defaults to &#x60;basic&#x60;.
          * @param {*} [options] Override http request option.
          * @throws {RequiredError}
          */
-        async testModel(id: string, options?: RawAxiosRequestConfig): Promise<(axios?: AxiosInstance, basePath?: string) => AxiosPromise<ModelTestResult>> {
-            const localVarAxiosArgs = await localVarAxiosParamCreator.testModel(id, options);
+        async testModel(id: string, mode?: ModelTestMode, options?: RawAxiosRequestConfig): Promise<(axios?: AxiosInstance, basePath?: string) => AxiosPromise<ModelTestResult>> {
+            const localVarAxiosArgs = await localVarAxiosParamCreator.testModel(id, mode, options);
             const localVarOperationServerIndex = configuration?.serverIndex ?? 0;
             const localVarOperationServerBasePath = operationServerMap['ModelsApi.testModel']?.[localVarOperationServerIndex]?.url;
             return (axios, basePath) => createRequestFunction(localVarAxiosArgs, globalAxios, BASE_PATH, configuration)(axios, localVarOperationServerBasePath || basePath);
@@ -168,23 +188,26 @@ export const ModelsApiFactory = function (configuration?: Configuration, basePat
         /**
          * Returns a paginated list of available LLM models from the built-in catalog.
          * @summary List Models
+         * @param {string} [provider] Filter models by provider name. Invalid values return &#x60;400&#x60;.
+         * @param {string} [query] Case-insensitive substring search across &#x60;id&#x60;, &#x60;display_name&#x60;, and &#x60;aliases&#x60;.
          * @param {number} [pageLimit] Maximum number of items to return per page.
          * @param {number} [pageOffset] Number of items to skip before returning results.
          * @param {*} [options] Override http request option.
          * @throws {RequiredError}
          */
-        listModels(pageLimit?: number, pageOffset?: number, options?: RawAxiosRequestConfig): AxiosPromise<PaginatedModelList> {
-            return localVarFp.listModels(pageLimit, pageOffset, options).then((request) => request(axios, basePath));
+        listModels(provider?: string, query?: string, pageLimit?: number, pageOffset?: number, options?: RawAxiosRequestConfig): AxiosPromise<PaginatedModelList> {
+            return localVarFp.listModels(provider, query, pageLimit, pageOffset, options).then((request) => request(axios, basePath));
         },
         /**
          * Tests a model by sending a simple prompt and reporting pass/fail.
          * @summary Test Model
          * @param {string} id The model identifier.
+         * @param {ModelTestMode} [mode] Test mode for the single-model test endpoint. Defaults to &#x60;basic&#x60;.
          * @param {*} [options] Override http request option.
          * @throws {RequiredError}
          */
-        testModel(id: string, options?: RawAxiosRequestConfig): AxiosPromise<ModelTestResult> {
-            return localVarFp.testModel(id, options).then((request) => request(axios, basePath));
+        testModel(id: string, mode?: ModelTestMode, options?: RawAxiosRequestConfig): AxiosPromise<ModelTestResult> {
+            return localVarFp.testModel(id, mode, options).then((request) => request(axios, basePath));
         },
     };
 };
@@ -196,24 +219,27 @@ export class ModelsApi extends BaseAPI {
     /**
      * Returns a paginated list of available LLM models from the built-in catalog.
      * @summary List Models
+     * @param {string} [provider] Filter models by provider name. Invalid values return &#x60;400&#x60;.
+     * @param {string} [query] Case-insensitive substring search across &#x60;id&#x60;, &#x60;display_name&#x60;, and &#x60;aliases&#x60;.
      * @param {number} [pageLimit] Maximum number of items to return per page.
      * @param {number} [pageOffset] Number of items to skip before returning results.
      * @param {*} [options] Override http request option.
      * @throws {RequiredError}
      */
-    public listModels(pageLimit?: number, pageOffset?: number, options?: RawAxiosRequestConfig) {
-        return ModelsApiFp(this.configuration).listModels(pageLimit, pageOffset, options).then((request) => request(this.axios, this.basePath));
+    public listModels(provider?: string, query?: string, pageLimit?: number, pageOffset?: number, options?: RawAxiosRequestConfig) {
+        return ModelsApiFp(this.configuration).listModels(provider, query, pageLimit, pageOffset, options).then((request) => request(this.axios, this.basePath));
     }
 
     /**
      * Tests a model by sending a simple prompt and reporting pass/fail.
      * @summary Test Model
      * @param {string} id The model identifier.
+     * @param {ModelTestMode} [mode] Test mode for the single-model test endpoint. Defaults to &#x60;basic&#x60;.
      * @param {*} [options] Override http request option.
      * @throws {RequiredError}
      */
-    public testModel(id: string, options?: RawAxiosRequestConfig) {
-        return ModelsApiFp(this.configuration).testModel(id, options).then((request) => request(this.axios, this.basePath));
+    public testModel(id: string, mode?: ModelTestMode, options?: RawAxiosRequestConfig) {
+        return ModelsApiFp(this.configuration).testModel(id, mode, options).then((request) => request(this.axios, this.basePath));
     }
 }
 
diff --git a/lib/packages/fabro-api-client/src/models/index.ts b/lib/packages/fabro-api-client/src/models/index.ts
index d61b98611..12e439975 100644
--- a/lib/packages/fabro-api-client/src/models/index.ts
+++ b/lib/packages/fabro-api-client/src/models/index.ts
@@ -67,6 +67,7 @@ export * from './model-costs';
 export * from './model-features';
 export * from './model-limits';
 export * from './model-reference';
+export * from './model-test-mode';
 export * from './model-test-result';
 export * from './node-state';
 export * from './node-status-record';
diff --git a/lib/packages/fabro-api-client/src/models/model-test-mode.ts b/lib/packages/fabro-api-client/src/models/model-test-mode.ts
new file mode 100644
index 000000000..853ff5913
--- /dev/null
+++ b/lib/packages/fabro-api-client/src/models/model-test-mode.ts
@@ -0,0 +1,29 @@
+/* tslint:disable */
+/* eslint-disable */
+/**
+ * Fabro Run API
+ * HTTP API for managing Fabro workflow run executions.
+ *
+ * The version of the OpenAPI document: 0.1.0
+ * 
+ *
+ * NOTE: This class is auto generated by OpenAPI Generator (https://openapi-generator.tech).
+ * https://openapi-generator.tech
+ * Do not edit the class manually.
+ */
+
+
+
+/**
+ * Single-model test mode.
+ */
+
+export const ModelTestMode = {
+    BASIC: 'basic',
+    DEEP: 'deep'
+} as const;
+
+export type ModelTestMode = typeof ModelTestMode[keyof typeof ModelTestMode];
+
+
+
diff --git a/lib/packages/fabro-api-client/src/models/model-test-result.ts b/lib/packages/fabro-api-client/src/models/model-test-result.ts
index 844d0c3db..8a1acfdbb 100644
--- a/lib/packages/fabro-api-client/src/models/model-test-result.ts
+++ b/lib/packages/fabro-api-client/src/models/model-test-result.ts
@@ -15,7 +15,7 @@
 
 
 /**
- * Result of testing a model with a simple prompt.
+ * Result of testing a model in `basic` or `deep` mode.
  */
 export interface ModelTestResult {
     /**