From 1e1c23107685efaa2a899274d02c2cc9d1014196 Mon Sep 17 00:00:00 2001 From: mateo-berri <277851410+mateo-berri@users.noreply.github.com> Date: Wed, 26 Aug 2026 16:52:09 -0700 Subject: [PATCH] fix(model_prices): bill gemini -latest/preview alias cache reads at 10% of input --- ...odel_prices_and_context_window_backup.json | 12 ++--- model_prices_and_context_window.json | 12 ++--- .../llms/gemini/test_cost_calculator.py | 49 +++++++++++++++++++ 3 files changed, 61 insertions(+), 12 deletions(-) diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index ad19818a5ce..7e3e1005bc0 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -20332,7 +20332,7 @@ "supports_image_size": false }, "gemini-2.5-flash-preview-09-2025": { - "cache_read_input_token_cost": 7.5e-08, + "cache_read_input_token_cost": 3e-08, "input_cost_per_audio_token": 1e-06, "input_cost_per_token": 3e-07, "litellm_provider": "vertex_ai-language-models", @@ -20469,7 +20469,7 @@ }, "gemini-2.5-flash-lite-preview-06-17": { "deprecation_date": "2025-11-18", - "cache_read_input_token_cost": 2.5e-08, + "cache_read_input_token_cost": 1e-08, "input_cost_per_audio_token": 5e-07, "input_cost_per_token": 1e-07, "litellm_provider": "vertex_ai-language-models", @@ -22062,7 +22062,7 @@ "supports_image_size": false }, "gemini/gemini-2.5-flash-preview-09-2025": { - "cache_read_input_token_cost": 7.5e-08, + "cache_read_input_token_cost": 3e-08, "deprecation_date": "2026-02-17", "input_cost_per_audio_token": 1e-06, "input_cost_per_token": 3e-07, @@ -22111,7 +22111,7 @@ "supports_image_size": false }, "gemini/gemini-flash-latest": { - "cache_read_input_token_cost": 7.5e-08, + "cache_read_input_token_cost": 3e-08, "input_cost_per_audio_token": 1e-06, "input_cost_per_token": 3e-07, "litellm_provider": "gemini", @@ -22158,7 +22158,7 @@ "google_maps_grounding_cost_per_query": 0.025 }, "gemini/gemini-flash-lite-latest": { - "cache_read_input_token_cost": 2.5e-08, + "cache_read_input_token_cost": 1e-08, "input_cost_per_audio_token": 3e-07, "input_cost_per_token": 1e-07, "litellm_provider": "gemini", @@ -22206,7 +22206,7 @@ }, "gemini/gemini-2.5-flash-lite-preview-06-17": { "deprecation_date": "2025-11-18", - "cache_read_input_token_cost": 2.5e-08, + "cache_read_input_token_cost": 1e-08, "input_cost_per_audio_token": 5e-07, "input_cost_per_token": 1e-07, "litellm_provider": "gemini", diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index ad19818a5ce..7e3e1005bc0 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -20332,7 +20332,7 @@ "supports_image_size": false }, "gemini-2.5-flash-preview-09-2025": { - "cache_read_input_token_cost": 7.5e-08, + "cache_read_input_token_cost": 3e-08, "input_cost_per_audio_token": 1e-06, "input_cost_per_token": 3e-07, "litellm_provider": "vertex_ai-language-models", @@ -20469,7 +20469,7 @@ }, "gemini-2.5-flash-lite-preview-06-17": { "deprecation_date": "2025-11-18", - "cache_read_input_token_cost": 2.5e-08, + "cache_read_input_token_cost": 1e-08, "input_cost_per_audio_token": 5e-07, "input_cost_per_token": 1e-07, "litellm_provider": "vertex_ai-language-models", @@ -22062,7 +22062,7 @@ "supports_image_size": false }, "gemini/gemini-2.5-flash-preview-09-2025": { - "cache_read_input_token_cost": 7.5e-08, + "cache_read_input_token_cost": 3e-08, "deprecation_date": "2026-02-17", "input_cost_per_audio_token": 1e-06, "input_cost_per_token": 3e-07, @@ -22111,7 +22111,7 @@ "supports_image_size": false }, "gemini/gemini-flash-latest": { - "cache_read_input_token_cost": 7.5e-08, + "cache_read_input_token_cost": 3e-08, "input_cost_per_audio_token": 1e-06, "input_cost_per_token": 3e-07, "litellm_provider": "gemini", @@ -22158,7 +22158,7 @@ "google_maps_grounding_cost_per_query": 0.025 }, "gemini/gemini-flash-lite-latest": { - "cache_read_input_token_cost": 2.5e-08, + "cache_read_input_token_cost": 1e-08, "input_cost_per_audio_token": 3e-07, "input_cost_per_token": 1e-07, "litellm_provider": "gemini", @@ -22206,7 +22206,7 @@ }, "gemini/gemini-2.5-flash-lite-preview-06-17": { "deprecation_date": "2025-11-18", - "cache_read_input_token_cost": 2.5e-08, + "cache_read_input_token_cost": 1e-08, "input_cost_per_audio_token": 5e-07, "input_cost_per_token": 1e-07, "litellm_provider": "gemini", diff --git a/tests/test_litellm/llms/gemini/test_cost_calculator.py b/tests/test_litellm/llms/gemini/test_cost_calculator.py index ba503d635a0..2f6f9a556c0 100644 --- a/tests/test_litellm/llms/gemini/test_cost_calculator.py +++ b/tests/test_litellm/llms/gemini/test_cost_calculator.py @@ -361,3 +361,52 @@ def test_gemini_image_generation_cost_no_web_search_when_absent(monkeypatch): ) assert cost_zero == cost_none + + +@pytest.mark.parametrize( + "model,custom_llm_provider,expected_cache_read_cost", + [ + ("gemini/gemini-flash-latest", "gemini", 3e-08), + ("gemini/gemini-flash-lite-latest", "gemini", 1e-08), + ("gemini/gemini-2.5-flash-preview-09-2025", "gemini", 3e-08), + ("gemini/gemini-2.5-flash-lite-preview-06-17", "gemini", 1e-08), + ("vertex_ai/gemini-2.5-flash-preview-09-2025", "vertex_ai", 3e-08), + ("vertex_ai/gemini-2.5-flash-lite-preview-06-17", "vertex_ai", 1e-08), + ], +) +def test_flash_alias_cache_read_is_ten_percent_of_input( + monkeypatch, model, custom_llm_provider, expected_cache_read_cost +): + monkeypatch.setenv("LITELLM_LOCAL_MODEL_COST_MAP", "True") + litellm.model_cost = litellm.get_model_cost_map(url="") + + model_info = litellm.get_model_info( + model=model, custom_llm_provider=custom_llm_provider + ) + + assert model_info["cache_read_input_token_cost"] == expected_cache_read_cost + assert model_info["cache_read_input_token_cost"] == pytest.approx( + 0.10 * model_info["input_cost_per_token"] + ) + + +@pytest.mark.parametrize( + "prefixed,bare", + [ + ("gemini/gemini-flash-latest", "gemini-flash-latest"), + ("gemini/gemini-flash-lite-latest", "gemini-flash-lite-latest"), + ], +) +def test_flash_latest_alias_spellings_price_identically(monkeypatch, prefixed, bare): + monkeypatch.setenv("LITELLM_LOCAL_MODEL_COST_MAP", "True") + litellm.model_cost = litellm.get_model_cost_map(url="") + + prefixed_entry = litellm.model_cost[prefixed] + bare_entry = litellm.model_cost[bare] + + for cost_key in ( + "input_cost_per_token", + "output_cost_per_token", + "cache_read_input_token_cost", + ): + assert prefixed_entry[cost_key] == bare_entry[cost_key]