fix(bedrock): price claude_platform calls from their own cost map rows (#48150)

Backport of #48150 to lts/1.108.x.
Cherry-picked from 560372456e (main).

litellm/llms/bedrock/cost_calculation.py

@@ -3,9 +3,13 @@ Helper util for handling bedrock-specific cost calculation
 - e.g.: prompt caching
 """
 
-from typing import TYPE_CHECKING
+from typing import TYPE_CHECKING, Final
 
-from litellm.litellm_core_utils.llm_cost_calc.utils import generic_cost_per_token
+import litellm
+from litellm.litellm_core_utils.llm_cost_calc.utils import (
+    generic_cost_per_token,
+    get_provider_specific_geo_multiplier,
+)
 
 if TYPE_CHECKING:
     from litellm.types.utils import Usage
@@ -17,9 +21,13 @@ def cost_per_token(model: str, usage: "Usage", service_tier: str | None = None)
 
     Follows the same logic as Anthropic's cost per token calculation.
     """
-    return generic_cost_per_token(
+    model_info: Final = litellm.get_model_info(model=model, custom_llm_provider="bedrock")
+    prompt_cost, completion_cost = generic_cost_per_token(
         model=model,
         usage=usage,
         custom_llm_provider="bedrock",
         service_tier=service_tier,
+        model_info=model_info,
     )
+    geo_multiplier: Final = get_provider_specific_geo_multiplier(model_info=model_info, usage=usage)
+    return prompt_cost * geo_multiplier, completion_cost * geo_multiplier

tests/unit/llms/bedrock/test_cost_calculation.py

@@ -0,0 +1,41 @@
+import json
+from importlib.resources import files
+from typing import Final
+
+import pytest
+
+import litellm
+from litellm.types.utils import Usage
+
+CLAUDE_PLATFORM_PREFIX: Final = "bedrock/claude_platform/"
+CLAUDE_PLATFORM_KEYS: Final = tuple(
+    key
+    for key in json.loads(files("litellm").joinpath("model_prices_and_context_window_backup.json").read_text())
+    if key.startswith(CLAUDE_PLATFORM_PREFIX)
+)
+
+
+def test_claude_platform_rows_exist() -> None:
+    assert CLAUDE_PLATFORM_KEYS
+
+
+@pytest.mark.parametrize("inference_geo", ["global", "us"])
+@pytest.mark.parametrize("key", CLAUDE_PLATFORM_KEYS)
+def test_claude_platform_cost_matches_claude_api_cost(local_model_cost_map: None, key: str, inference_geo: str) -> None:
+    usage: Final = Usage(
+        prompt_tokens=1_000,
+        completion_tokens=200,
+        cache_read_input_tokens=300,
+        cache_creation_input_tokens=250,
+        inference_geo=inference_geo,
+    )
+
+    claude_platform_cost: Final = litellm.cost_per_token(
+        model=key.removeprefix("bedrock/"), custom_llm_provider="bedrock", usage_object=usage
+    )
+    claude_api_cost: Final = litellm.cost_per_token(
+        model=key.removeprefix(CLAUDE_PLATFORM_PREFIX), custom_llm_provider="anthropic", usage_object=usage
+    )
+
+    assert claude_platform_cost == pytest.approx(claude_api_cost)
+    assert sum(claude_platform_cost) > 0

tests/unit/test_utils_get_model_info.py

@@ -136,6 +136,7 @@ def _enforce_bedrock_converse_models(
             info.get("litellm_provider") == "bedrock"
             and info.get("mode") == "chat"
             and model not in whitelist_models
+            and BedrockModelInfo.get_bedrock_route(model) != "claude_platform"
             and not (
                 (base_model := BedrockModelInfo.get_base_model(model)) != model
                 and model_cost.get(base_model, {}).get("litellm_provider") == "bedrock_converse"