diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index 02df6897b41..30b08e54410 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -50868,7 +50868,7 @@ "vertex_ai/google/gemma-4-26b-a4b-it-maas": { "input_cost_per_token": 1.5e-07, "litellm_provider": "vertex_ai-openai_models", - "max_input_tokens": 256000, + "max_input_tokens": 262144, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index 02df6897b41..30b08e54410 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -50868,7 +50868,7 @@ "vertex_ai/google/gemma-4-26b-a4b-it-maas": { "input_cost_per_token": 1.5e-07, "litellm_provider": "vertex_ai-openai_models", - "max_input_tokens": 256000, + "max_input_tokens": 262144, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", diff --git a/tests/test_litellm/llms/vertex_ai/vertex_ai_partner_models/gemma/test_vertex_ai_gemma_global_endpoint.py b/tests/test_litellm/llms/vertex_ai/vertex_ai_partner_models/gemma/test_vertex_ai_gemma_global_endpoint.py index e9b58622a4b..85a2124ab02 100644 --- a/tests/test_litellm/llms/vertex_ai/vertex_ai_partner_models/gemma/test_vertex_ai_gemma_global_endpoint.py +++ b/tests/test_litellm/llms/vertex_ai/vertex_ai_partner_models/gemma/test_vertex_ai_gemma_global_endpoint.py @@ -30,7 +30,7 @@ from litellm.types.llms.vertex_ai import VertexPartnerProvider _GEMMA_MODEL_COST_ENTRY = { "vertex_ai/google/gemma-4-26b-a4b-it-maas": { "litellm_provider": "vertex_ai-openai_models", - "max_input_tokens": 256000, + "max_input_tokens": 262144, "max_output_tokens": 128000, "max_tokens": 128000, "mode": "chat", @@ -180,6 +180,15 @@ class TestCreateVertexURLGemma: # --------------------------------------------------------------------------- +def test_gemma_maas_context_window_matches_google(local_model_cost_map): + info = litellm.get_model_info("vertex_ai/google/gemma-4-26b-a4b-it-maas") + + # 262,144 context length and 128,000 maximum output per Google's model page, checked 2026-09-18: + # https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/maas/google/gemma-4-26b-a4b-it + assert info["max_input_tokens"] == 262144 + assert info["max_output_tokens"] == 128000 + + # --------------------------------------------------------------------------- # Integration tests: verify payloads reach the global OpenAI endpoint #