fix(cost_map): fill missing Claude capability fields from the family baseline

Mapped Claude entries on Anthropic, Azure Foundry, Bedrock and Vertex now get
supports_parallel_function_calling and supports_system_messages from the
claude-family-baseline rule when the entry leaves them out. Values set on an
entry still win

litellm/model_prices_and_context_window_backup.json

@@ -63749,7 +63749,8 @@
             {
                 "name": "claude-family-baseline",
                 "pattern": "claude-[a-z]+-\\d+(?:[-.]\\d+)?(?:-\\d{8})?",
-                "description": "Any Claude family-major id with an optional minor and an optional 8-digit date suffix, under any provider namespace (bare, bedrock-dotted, vertex, databricks, ...), so bare majors like claude-newfamily-5 get the same baseline as claude-newfamily-5-1. Carries the model-family facts every Claude shares; it carries no pricing, so cost stays on the standard unpriced behavior rather than a guessed number.",
+                "fill_missing_for_providers": ["anthropic", "azure_ai", "bedrock", "bedrock_converse", "vertex_ai-anthropic_models"],
+                "description": "Any Claude family-major id with an optional minor and an optional 8-digit date suffix, under any provider namespace (bare, bedrock-dotted, vertex, databricks, ...), so bare majors like claude-newfamily-5 get the same baseline as claude-newfamily-5-1. Carries the model-family facts every Claude shares; it carries no pricing, so cost stays on the standard unpriced behavior rather than a guessed number. On mapped Claude entries from Anthropic, Azure Foundry, Bedrock and Vertex it only fills the keys an entry leaves out, so the entry's own values always win.",
                 "model_info": {
                     "mode": "chat",
                     "max_input_tokens": 200000,

model_prices_and_context_window.json

@@ -63749,7 +63749,8 @@
             {
                 "name": "claude-family-baseline",
                 "pattern": "claude-[a-z]+-\\d+(?:[-.]\\d+)?(?:-\\d{8})?",
-                "description": "Any Claude family-major id with an optional minor and an optional 8-digit date suffix, under any provider namespace (bare, bedrock-dotted, vertex, databricks, ...), so bare majors like claude-newfamily-5 get the same baseline as claude-newfamily-5-1. Carries the model-family facts every Claude shares; it carries no pricing, so cost stays on the standard unpriced behavior rather than a guessed number.",
+                "fill_missing_for_providers": ["anthropic", "azure_ai", "bedrock", "bedrock_converse", "vertex_ai-anthropic_models"],
+                "description": "Any Claude family-major id with an optional minor and an optional 8-digit date suffix, under any provider namespace (bare, bedrock-dotted, vertex, databricks, ...), so bare majors like claude-newfamily-5 get the same baseline as claude-newfamily-5-1. Carries the model-family facts every Claude shares; it carries no pricing, so cost stays on the standard unpriced behavior rather than a guessed number. On mapped Claude entries from Anthropic, Azure Foundry, Bedrock and Vertex it only fills the keys an entry leaves out, so the entry's own values always win.",
                 "model_info": {
                     "mode": "chat",
                     "max_input_tokens": 200000,

tests/unit/litellm_core_utils/test_fallback_generalizations.py

@@ -9,7 +9,8 @@ resolution (get_model_info) including the shipped rules in the bundled cost map.
 
 import json
 import logging
-from collections.abc import Callable
+import re
+from collections.abc import Callable, Mapping
 from pathlib import Path
 from typing import Final
 
@@ -914,11 +915,11 @@ def test_shipped_claude_thinking_rules_backfill_only_anthropic(shipped_cost_map)
     assert info.get("supports_adaptive_thinking") is None
     assert info.get("supports_legacy_thinking") is None
     assert info.get("max_input_tokens") is None
-    assert match_fill_missing_generalizations("claude-sonnet-4-6", "anthropic") == {
+    assert match_fill_missing_generalizations("claude-sonnet-4-6", "anthropic").items() >= {
         "supports_adaptive_thinking": True,
         "supports_legacy_thinking": True,
         "supports_tool_search": True,
-    }
+    }.items()
     assert match_fill_missing_generalizations("claude-sonnet-4-6", "perplexity") is None
 
 
@@ -961,3 +962,55 @@ def test_shipped_tool_search_rule_fills_mapped_claude_entries_without_flag(shipp
     assert match_fill_missing_generalizations("claude-opus-5", "bedrock")["supports_tool_search"] is True
     assert match_fill_missing_generalizations("claude-opus-5", "azure_ai")["supports_tool_search"] is True
     assert match_fill_missing_generalizations("claude-opus-5", "perplexity") is None
+
+
+CLAUDE_BASELINE_FILL_PROVIDERS: Final = ("anthropic", "azure_ai", "bedrock", "bedrock_converse", "vertex_ai-anthropic_models")
+
+
+def _claude_baseline() -> Mapping[str, object]:
+    return next(rule for rule in get_fallback_generalization_rules() if rule["name"] == "claude-family-baseline")
+
+
+def _mapped_claude_chat_entries() -> tuple[tuple[str, Mapping[str, object]], ...]:
+    pattern: Final = re.compile(_claude_baseline()["pattern"], re.IGNORECASE)
+    return tuple(
+        (key, entry)
+        for key, entry in litellm.model_cost.items()
+        if isinstance(entry, dict) and entry.get("mode") == "chat" and pattern.search(key) is not None
+    )
+
+
+def _fields_overriding_baseline(entry: Mapping[str, object], baseline: Mapping[str, object]) -> frozenset[str]:
+    return frozenset(field for field in entry.keys() & baseline.keys() if entry[field] != baseline[field])
+
+
+@pytest.mark.parametrize("provider", CLAUDE_BASELINE_FILL_PROVIDERS)
+def test_shipped_claude_baseline_fills_only_missing_keys_on_mapped_entries(shipped_cost_map, provider):
+    """A mapped Claude entry from a first-party host gets each baseline key it leaves out,
+    while every baseline key the entry sets keeps the entry's own value."""
+    baseline: Final = _claude_baseline()["model_info"]
+    entries: Final = tuple((key, entry) for key, entry in _mapped_claude_chat_entries() if entry["litellm_provider"] == provider)
+    assert any(baseline.keys() - entry.keys() for _, entry in entries)
+    assert any(_fields_overriding_baseline(entry, baseline) for _, entry in entries)
+
+    for key, entry in entries:
+        info = litellm.get_model_info(key)
+        assert info["key"] == key
+        assert {field: info[field] for field in baseline} == {
+            field: entry.get(field, value) for field, value in baseline.items()
+        }, key
+
+
+def test_shipped_claude_baseline_leaves_other_providers_entries_unfilled(shipped_cost_map):
+    baseline: Final = _claude_baseline()["model_info"]
+    entries: Final = tuple(
+        (key, entry)
+        for key, entry in _mapped_claude_chat_entries()
+        if entry["litellm_provider"] not in CLAUDE_BASELINE_FILL_PROVIDERS
+    )
+    assert any("supports_parallel_function_calling" not in entry for _, entry in entries)
+
+    for key, entry in entries:
+        info = litellm.get_model_info(key)
+        missing = baseline.keys() - entry.keys()
+        assert {field: info.get(field) for field in missing} == dict.fromkeys(missing), key