feat(providers): add Pareto Inference as OpenAI-compatible provider

Registers pareto through the JSON provider registry, so pareto/z-ai/glm-5.3-flash
resolves to https://api.paretoinference.com/v1 with PARETO_API_KEY and an optional
PARETO_API_BASE override. Adds LlmProviders.PARETO, the constants.py provider and
endpoint lists so an api_base pointing at Pareto also resolves to the provider, the
provider endpoints matrix, the Add Model provider fields, per-token pricing for
pareto/z-ai/glm-5.3-flash in both cost maps, and tests.

litellm/constants.py

@@ -1095,6 +1095,7 @@ openai_compatible_endpoints: Final[list[str]] = [
     "https://api.prisminference.com/v1",
     "https://api.reka.ai/v1",
     "https://inference.runanywhere.ai/v1",
+    "https://api.paretoinference.com/v1",
     "https://gigachat.devices.sberbank.ru/api/v1",
     "https://api.openinfer.ai/v1",
     "https://vanchin.streamlake.ai/api/gateway/v1/endpoints",
@@ -1183,6 +1184,7 @@ openai_compatible_providers: Final[list[str]] = [
     "reka",
     "streamlake",
     "wally",
+    "pareto",
 ]
 
 OPENAI_AUDIO_TRANSCRIPTION_PROVIDERS: Final = frozenset({"openai"} | frozenset(openai_compatible_providers))

litellm/llms/openai_like/providers.json

@@ -314,5 +314,10 @@
       "supports_function_calling": true,
       "supports_reasoning": true
     }
+  },
+  "pareto": {
+    "base_url": "https://api.paretoinference.com/v1",
+    "api_key_env": "PARETO_API_KEY",
+    "api_base_env": "PARETO_API_BASE"
   }
 }

litellm/model_prices_and_context_window_backup.json

@@ -64400,6 +64400,14 @@
         "supports_system_messages": true,
         "supports_tool_choice": true
     },
+    "pareto/z-ai/glm-5.3-flash": {
+        "input_cost_per_token": 3e-08,
+        "output_cost_per_token": 1e-07,
+        "cache_read_input_token_cost": 6e-09,
+        "litellm_provider": "pareto",
+        "mode": "chat",
+        "source": "https://docs.paretoinference.com/pricing"
+    },
     "xai/grok-4.20-0309-non-reasoning": {
         "cache_read_input_token_cost": 2e-07,
         "cache_read_input_token_cost_batches": 1.6e-07,

litellm/provider_endpoints_support_backup.json

@@ -3189,6 +3189,23 @@
         "generateContent": false
       }
     },
+    "pareto": {
+      "display_name": "Pareto Inference (`pareto`)",
+      "url": "https://docs.litellm.ai/docs/providers/pareto",
+      "endpoints": {
+        "chat_completions": true,
+        "messages": false,
+        "responses": false,
+        "embeddings": false,
+        "image_generations": false,
+        "audio_transcriptions": false,
+        "audio_speech": false,
+        "moderations": false,
+        "batches": false,
+        "rerank": false,
+        "a2a": false
+      }
+    },
     "perplexity": {
       "display_name": "Perplexity AI (`perplexity`)",
       "url": "https://docs.litellm.ai/docs/providers/perplexity",

litellm/proxy/public_endpoints/provider_create_fields.json

@@ -3269,6 +3269,34 @@
     ],
     "default_model_placeholder": "gpt-3.5-turbo"
   },
+  {
+    "provider": "PARETO",
+    "provider_display_name": "Pareto Inference",
+    "litellm_provider": "pareto",
+    "credential_fields": [
+      {
+        "key": "api_base",
+        "label": "API Base",
+        "placeholder": "https://api.paretoinference.com/v1",
+        "tooltip": null,
+        "required": false,
+        "field_type": "text",
+        "options": null,
+        "default_value": null
+      },
+      {
+        "key": "api_key",
+        "label": "API Key",
+        "placeholder": null,
+        "tooltip": null,
+        "required": true,
+        "field_type": "password",
+        "options": null,
+        "default_value": null
+      }
+    ],
+    "default_model_placeholder": "pareto/z-ai/glm-5.3-flash"
+  },
   {
     "provider": "Perplexity",
     "provider_display_name": "Perplexity",

litellm/types/utils.py

@@ -4785,6 +4785,7 @@ class LlmProviders(str, Enum):
     INFERENCEAPIS = "inferenceapis"
     HOSTINGER_ROUTER = "hostinger_router"
     ZEROGPU = "zerogpu"
+    PARETO = "pareto"
     DARKBLOOM = "darkbloom"
     META = "meta"
     SAIL = "sail"

model_prices_and_context_window.json

@@ -64400,6 +64400,14 @@
         "supports_system_messages": true,
         "supports_tool_choice": true
     },
+    "pareto/z-ai/glm-5.3-flash": {
+        "input_cost_per_token": 3e-08,
+        "output_cost_per_token": 1e-07,
+        "cache_read_input_token_cost": 6e-09,
+        "litellm_provider": "pareto",
+        "mode": "chat",
+        "source": "https://docs.paretoinference.com/pricing"
+    },
     "xai/grok-4.20-0309-non-reasoning": {
         "cache_read_input_token_cost": 2e-07,
         "cache_read_input_token_cost_batches": 1.6e-07,

provider_endpoints_support.json

@@ -3189,6 +3189,23 @@
         "generateContent": false
       }
     },
+    "pareto": {
+      "display_name": "Pareto Inference (`pareto`)",
+      "url": "https://docs.litellm.ai/docs/providers/pareto",
+      "endpoints": {
+        "chat_completions": true,
+        "messages": false,
+        "responses": false,
+        "embeddings": false,
+        "image_generations": false,
+        "audio_transcriptions": false,
+        "audio_speech": false,
+        "moderations": false,
+        "batches": false,
+        "rerank": false,
+        "a2a": false
+      }
+    },
     "perplexity": {
       "display_name": "Perplexity AI (`perplexity`)",
       "url": "https://docs.litellm.ai/docs/providers/perplexity",

tests/unit/llms/openai_like/test_pareto_provider.py

@@ -0,0 +1,114 @@
+import json
+from pathlib import Path
+from typing import Final
+
+import pytest
+import respx
+
+import litellm
+from litellm.litellm_core_utils.get_llm_provider_logic import get_llm_provider
+from litellm.llms.openai_like.json_loader import JSONProviderRegistry, SimpleProviderConfig
+
+MODEL: Final = "z-ai/glm-5.3-flash"
+REPO_ROOT: Final = Path(__file__).parents[4]
+COST_MAPS: Final = (
+    REPO_ROOT / "model_prices_and_context_window.json",
+    REPO_ROOT / "litellm" / "model_prices_and_context_window_backup.json",
+)
+
+
+@pytest.fixture
+def pareto() -> SimpleProviderConfig:
+    config: Final = JSONProviderRegistry.get("pareto")
+    assert config is not None
+    return config
+
+
+def test_prefixed_model_resolves_to_pareto_and_keeps_the_vendor_namespace(
+    pareto: SimpleProviderConfig, monkeypatch: pytest.MonkeyPatch
+) -> None:
+    monkeypatch.delenv("PARETO_API_BASE", raising=False)
+
+    model, provider, _, api_base = get_llm_provider(model=f"pareto/{MODEL}")
+
+    assert provider == litellm.LlmProviders.PARETO.value
+    assert model == MODEL
+    assert api_base == pareto.base_url
+
+
+def test_api_key_and_base_are_read_from_pareto_env_vars(monkeypatch: pytest.MonkeyPatch) -> None:
+    monkeypatch.delenv("OPENAI_API_KEY", raising=False)
+    monkeypatch.setenv("PARETO_API_KEY", "sk-pareto-env")
+    monkeypatch.setenv("PARETO_API_BASE", "https://pareto.internal.example/v1")
+
+    _, provider, api_key, api_base = get_llm_provider(model=f"pareto/{MODEL}")
+
+    assert provider == "pareto"
+    assert api_key == "sk-pareto-env"
+    assert api_base == "https://pareto.internal.example/v1"
+
+
+def test_default_api_base_autodetects_pareto(pareto: SimpleProviderConfig, monkeypatch: pytest.MonkeyPatch) -> None:
+    monkeypatch.setenv("PARETO_API_KEY", "sk-pareto-env")
+
+    model, provider, api_key, _ = get_llm_provider(model=MODEL, api_base=pareto.base_url)
+
+    assert provider == "pareto"
+    assert model == MODEL
+    assert api_key == "sk-pareto-env"
+
+
+@respx.mock
+def test_completion_posts_to_pareto_chat_completions_with_bearer_key(
+    pareto: SimpleProviderConfig, monkeypatch: pytest.MonkeyPatch
+) -> None:
+    monkeypatch.delenv("PARETO_API_BASE", raising=False)
+    monkeypatch.setenv("PARETO_API_KEY", "sk-pareto-env")
+    route: Final = respx.post(f"{pareto.base_url}/chat/completions").respond(
+        json={
+            "id": "chatcmpl-pareto",
+            "object": "chat.completion",
+            "created": 1,
+            "model": MODEL,
+            "choices": [{"index": 0, "message": {"role": "assistant", "content": "hi"}, "finish_reason": "stop"}],
+            "usage": {"prompt_tokens": 5, "completion_tokens": 1, "total_tokens": 6},
+        }
+    )
+
+    response: Final = litellm.completion(
+        model=f"pareto/{MODEL}",
+        messages=[{"role": "user", "content": "hello"}],
+    )
+
+    assert route.call_count == 1
+    request: Final = route.calls.last.request
+    assert request.headers["authorization"] == "Bearer sk-pareto-env"
+    assert json.loads(request.content)["model"] == MODEL
+    assert response.choices[0].message.content == "hi"
+    assert response.usage.total_tokens == 6
+
+
+@pytest.fixture
+def local_model_cost_map(monkeypatch: pytest.MonkeyPatch):
+    """Resolve pricing against the in-repo cost map, not the remote one fetched at import."""
+    monkeypatch.setenv("LITELLM_LOCAL_MODEL_COST_MAP", "True")
+    monkeypatch.setattr(litellm, "model_cost", litellm.get_model_cost_map(url=""))
+    litellm.get_model_info.cache_clear()
+    yield
+    litellm.get_model_info.cache_clear()
+
+
+def test_cost_maps_carry_the_same_pareto_entry() -> None:
+    main, backup = (json.loads(path.read_text())[f"pareto/{MODEL}"] for path in COST_MAPS)
+
+    assert main == backup
+    assert main["litellm_provider"] == "pareto"
+
+
+def test_published_per_token_prices_are_used_for_cost(local_model_cost_map: None) -> None:
+    prompt_cost, completion_cost = litellm.cost_per_token(
+        model=f"pareto/{MODEL}", prompt_tokens=1_000_000, completion_tokens=1_000_000
+    )
+
+    assert prompt_cost == pytest.approx(0.03)
+    assert completion_cost == pytest.approx(0.10)