Fix Vertex Gemma 4 routing metadata

litellm/llms/vertex_ai/common_utils.py

@@ -50,7 +50,7 @@ def get_vertex_ai_model_route(
     Determine which handler to use for a Vertex AI model based on the model name.
 
     Args:
-        model: The model name (e.g., "llama3-405b", "gemini-pro", "gemma/gemma-3-12b-it", "openai/gpt-oss-120b")
+        model: The model name (e.g., "llama3-405b", "gemini-pro", "gemma/gemma-3-12b-it", "google/gemma-4-26b-a4b-it-maas", "openai/gpt-oss-120b")
         litellm_params: Optional litellm parameters dict that may contain base_model for routing
 
     Returns:
@@ -66,6 +66,9 @@ def get_vertex_ai_model_route(
         >>> get_vertex_ai_model_route("gemma/gemma-3-12b-it")
         VertexAIModelRoute.GEMMA
 
+        >>> get_vertex_ai_model_route("google/gemma-4-26b-a4b-it-maas")
+        VertexAIModelRoute.GEMMA
+
         >>> get_vertex_ai_model_route("openai/gpt-oss-120b")
         VertexAIModelRoute.MODEL_GARDEN
 
@@ -99,7 +102,7 @@ def get_vertex_ai_model_route(
         return VertexAIModelRoute.BGE
 
     # Check for gemma models
-    if "gemma/" in model:
+    if "gemma/" in model or model.startswith("google/gemma-"):
         return VertexAIModelRoute.GEMMA
 
     # Check for model garden openai models

litellm/model_prices_and_context_window_backup.json

@@ -32735,14 +32735,15 @@
     "vertex_ai/google/gemma-4-26b-a4b-it-maas": {
         "input_cost_per_token": 1.5e-07,
         "litellm_provider": "vertex_ai",
-        "max_input_tokens": 256000,
-        "max_output_tokens": 8192,
-        "max_tokens": 8192,
+        "max_input_tokens": 262144,
+        "max_output_tokens": 128000,
+        "max_tokens": 128000,
         "mode": "chat",
         "output_cost_per_token": 6e-07,
         "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing",
         "supports_function_calling": true,
         "supports_system_messages": true,
+        "supports_tool_choice": true,
         "supports_vision": true
     },
     "vertex_ai/deep-research-pro-preview-12-2025": {

model_prices_and_context_window.json

@@ -32789,14 +32789,15 @@
     "vertex_ai/google/gemma-4-26b-a4b-it-maas": {
         "input_cost_per_token": 1.5e-07,
         "litellm_provider": "vertex_ai",
-        "max_input_tokens": 256000,
-        "max_output_tokens": 8192,
-        "max_tokens": 8192,
+        "max_input_tokens": 262144,
+        "max_output_tokens": 128000,
+        "max_tokens": 128000,
         "mode": "chat",
         "output_cost_per_token": 6e-07,
         "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing",
         "supports_function_calling": true,
         "supports_system_messages": true,
+        "supports_tool_choice": true,
         "supports_vision": true
     },
     "vertex_ai/deep-research-pro-preview-12-2025": {

tests/test_litellm/llms/vertex_ai/vertex_gemma_models/test_vertex_gemma_transformation.py

@@ -5,6 +5,7 @@ Maps to: litellm/llms/vertex_ai/vertex_gemma_models/transformation.py
 """
 
 import json
+from pathlib import Path
 from unittest.mock import AsyncMock, Mock, patch
 
 import pytest
@@ -20,6 +21,40 @@ def _reset_litellm_http_client_cache():
     in_memory_llm_clients_cache.flush_cache()
 
 
+def test_gemma_4_26b_pricing_metadata_uses_vertex_gemma_route():
+    from litellm.llms.vertex_ai.common_utils import (
+        VertexAIModelRoute,
+        get_vertex_ai_model_route,
+    )
+
+    model = "google/gemma-4-26b-a4b-it-maas"
+    litellm_model = f"vertex_ai/{model}"
+
+    assert get_vertex_ai_model_route(model=model) == VertexAIModelRoute.GEMMA
+    assert (
+        get_vertex_ai_model_route(model="gemma/gemma-3-12b-it")
+        == VertexAIModelRoute.GEMMA
+    )
+
+    repo_root = Path(litellm.__file__).resolve().parents[1]
+    model_cost_files = [
+        repo_root / "model_prices_and_context_window.json",
+        repo_root / "litellm" / "model_prices_and_context_window_backup.json",
+    ]
+
+    for model_cost_file in model_cost_files:
+        model_cost = json.loads(model_cost_file.read_text(encoding="utf-8"))
+
+        assert "vertex_ai/gemma/gemma-4-26b-a4b-it-maas" not in model_cost
+        model_info = model_cost[litellm_model]
+
+        assert model_info["max_input_tokens"] == 262144
+        assert model_info["max_output_tokens"] == 128000
+        assert model_info["max_tokens"] == 128000
+        assert model_info["supports_function_calling"] is True
+        assert model_info["supports_tool_choice"] is True
+
+
 class TestVertexGemmaCompletion:
     """Test completion flow for Vertex AI Gemma models using litellm.acompletion()"""