From 0243c5dee487f37cd7e9e1f7902bd1353e34eff6 Mon Sep 17 00:00:00 2001
From: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
Date: Wed, 26 Aug 2026 16:25:15 -0700
Subject: [PATCH] fix(model_prices): correct gemini-3.5-flash-lite flex
 cache-read pricing

---
 ...odel_prices_and_context_window_backup.json |  6 +--
 model_prices_and_context_window.json          |  6 +--
 .../llm_cost_calc/test_llm_cost_calc_utils.py | 38 +++++++++++++++++++
 3 files changed, 44 insertions(+), 6 deletions(-)

diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json
index a6da2c1fb09..cead28e18aa 100644
--- a/litellm/model_prices_and_context_window_backup.json
+++ b/litellm/model_prices_and_context_window_backup.json
@@ -20147,7 +20147,7 @@
     "gemini-3.5-flash-lite": {
         "deprecation_date": "2027-07-21",
         "cache_read_input_token_cost": 3e-08,
-        "cache_read_input_token_cost_flex": 2e-08,
+        "cache_read_input_token_cost_flex": 1.5e-08,
         "cache_read_input_token_cost_priority": 5e-08,
         "input_cost_per_token": 3e-07,
         "input_cost_per_token_batches": 1.5e-07,
@@ -22488,7 +22488,7 @@
     },
     "gemini/gemini-3.5-flash-lite": {
         "cache_read_input_token_cost": 3e-08,
-        "cache_read_input_token_cost_flex": 2e-08,
+        "cache_read_input_token_cost_flex": 1.5e-08,
         "cache_read_input_token_cost_priority": 5e-08,
         "input_cost_per_token": 3e-07,
         "input_cost_per_token_batches": 1.5e-07,
@@ -41987,7 +41987,7 @@
     "vertex_ai/gemini-3.5-flash-lite": {
         "deprecation_date": "2027-07-21",
         "cache_read_input_token_cost": 3e-08,
-        "cache_read_input_token_cost_flex": 2e-08,
+        "cache_read_input_token_cost_flex": 1.5e-08,
         "cache_read_input_token_cost_priority": 5e-08,
         "input_cost_per_token": 3e-07,
         "input_cost_per_token_batches": 1.5e-07,
diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json
index a6da2c1fb09..cead28e18aa 100644
--- a/model_prices_and_context_window.json
+++ b/model_prices_and_context_window.json
@@ -20147,7 +20147,7 @@
     "gemini-3.5-flash-lite": {
         "deprecation_date": "2027-07-21",
         "cache_read_input_token_cost": 3e-08,
-        "cache_read_input_token_cost_flex": 2e-08,
+        "cache_read_input_token_cost_flex": 1.5e-08,
         "cache_read_input_token_cost_priority": 5e-08,
         "input_cost_per_token": 3e-07,
         "input_cost_per_token_batches": 1.5e-07,
@@ -22488,7 +22488,7 @@
     },
     "gemini/gemini-3.5-flash-lite": {
         "cache_read_input_token_cost": 3e-08,
-        "cache_read_input_token_cost_flex": 2e-08,
+        "cache_read_input_token_cost_flex": 1.5e-08,
         "cache_read_input_token_cost_priority": 5e-08,
         "input_cost_per_token": 3e-07,
         "input_cost_per_token_batches": 1.5e-07,
@@ -41987,7 +41987,7 @@
     "vertex_ai/gemini-3.5-flash-lite": {
         "deprecation_date": "2027-07-21",
         "cache_read_input_token_cost": 3e-08,
-        "cache_read_input_token_cost_flex": 2e-08,
+        "cache_read_input_token_cost_flex": 1.5e-08,
         "cache_read_input_token_cost_priority": 5e-08,
         "input_cost_per_token": 3e-07,
         "input_cost_per_token_batches": 1.5e-07,
diff --git a/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py b/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py
index e13643ed6ce..18c3014c35d 100644
--- a/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py
+++ b/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py
@@ -3377,6 +3377,44 @@ def test_generic_cost_per_token_gemini_35_flash_lite(_local_model_cost_map):
     assert completion_cost == pytest.approx(0.00125)
 
 
+GEMINI_35_FLASH_LITE_SERVICE_TIER_PRICING = [
+    (None, 3e-07, 2.5e-06, 3e-08),
+    ("flex", 1.5e-07, 1.25e-06, 1.5e-08),
+    ("priority", 5.4e-07, 4.5e-06, 5e-08),
+]
+
+
+@pytest.mark.parametrize(
+    "service_tier,input_rate,output_rate,cache_read_rate", GEMINI_35_FLASH_LITE_SERVICE_TIER_PRICING
+)
+@pytest.mark.parametrize(
+    "model",
+    ["gemini-3.5-flash-lite", "gemini/gemini-3.5-flash-lite", "vertex_ai/gemini-3.5-flash-lite"],
+)
+def test_gemini_35_flash_lite_service_tier_pricing(
+    model, service_tier, input_rate, output_rate, cache_read_rate, _local_model_cost_map
+):
+    """Regression: Vertex publishes flash-lite Flex/Batch context caching at $0.015/M
+    (1.5e-08/token), so flex cache reads must not be billed at the 2e-08 rate the map
+    used to carry."""
+    usage = Usage(
+        prompt_tokens=1_000,
+        completion_tokens=500,
+        total_tokens=1_500,
+        prompt_tokens_details=PromptTokensDetailsWrapper(cached_tokens=200, text_tokens=800),
+    )
+
+    prompt_cost, completion_cost = generic_cost_per_token(
+        model=model.split("/")[-1],
+        usage=usage,
+        custom_llm_provider=model.split("/")[0] if "/" in model else "gemini",
+        service_tier=service_tier,
+    )
+
+    assert prompt_cost == pytest.approx(800 * input_rate + 200 * cache_read_rate, rel=1e-9)
+    assert completion_cost == pytest.approx(500 * output_rate, rel=1e-9)
+
+
 @pytest.mark.parametrize(
     "service_tier,input_rate,cache_read_rate,cache_write_rate,output_rate",
     [