fix(cost_map): vertex gemini-3.5-flash-lite priority cache read is $0.054/M

[devin-ai-integration]

TLDR

Problem this solves:

• Vertex priority cached input for gemini-3.5-flash-lite is $0.054/M, map carried $0.05/M
• test_gemini_35_flash_lite_service_tier_pricing hardcodes the stale $0.05/M and fails on the pricing sync PR chore(prices): sync Vertex AI prices: 14 models #40955

How it solves it:

• Set cache_read_input_token_cost_priority to 5.4e-08 on gemini-3.5-flash-lite and vertex_ai/gemini-3.5-flash-lite in both map copies
• Update the vertex_ai priority row in GEMINI_35_FLASH_LITE_TIER_RATES_BY_SURFACE to 5.4e-08
• gemini/gemini-3.5-flash-lite (Gemini API surface) is untouched at 5e-08

User Flow

Before: a Vertex user on the priority tier with cached prompt tokens is underbilled

1. They send POST https://litellm-domain/v1/chat/completions with "model": "vertex_ai/gemini-3.5-flash-lite", "service_tier": "priority" and a prompt that hits the context cache
2. The response header x-litellm-response-cost bills cached input at $0.05/M
3. https://litellm-domain/ui/?page=logs shows spend about 7% below the Google invoice for cached input

After: the same request bills cached input at Google's published priority rate

1. They send the same POST https://litellm-domain/v1/chat/completions
2. x-litellm-response-cost bills cached input at $0.054/M
3. https://litellm-domain/ui/?page=logs matches the Google invoice

Relevant issues

Unblocks CI on #40955 (core-utils shard). The provider sync produced the correct 5.4e-08 there; the hardcoded test was stale

Affected release

Linear ticket

Pre-Submission checklist

• I have added meaningful tests
• The handful of test files covering my change pass locally, e.g. uv run pytest tests/test_litellm/<your_test_file>.py -v
• My PR passes all required CI/CD checks
• My PR's scope is as isolated as possible; it only solves 1 specific problem
• I have received a Greptile Confidence Score of at least 4/5 before requesting a maintainer review

Screenshots / Proof of Fix

Raw page evidence, fetched with curl and a cache-buster from https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing, Gemini 3 section, Priority tab, columns "Price (/1M tokens) <= 200K input tokens with Priority | > 200K input with Priority | <= 200K cached input tokens with Priority | > 200K cached input tokens with Priority":

Gemini 3.5 Flash-Lite | Input (text, image, video, audio) | Global       | $0.54  | $0.54  | $0.054  | $0.054
                      |                                   | Non-global * | $0.594 | $0.594 | $0.0594 | $0.0594
                      | Text output (response and reasoning) | Global    | $4.50  | $4.50  | N/A     | N/A

$0.054 per 1M tokens = 5.4e-08 per token, matching the new map value. Input $0.54/M (5.4e-07) and output $4.50/M (4.5e-06) already matched

Before (7fd541e)

1. LITELLM_LOCAL_MODEL_COST_MAP=True uv run pytest tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py -k gemini_35_flash_lite with the chore(prices): sync Vertex AI prices: 14 models #40955 map: FAILED ...[vertex_ai-priority-5.4e-07-4.5e-06-5e-08] - assert 0.0004428 == 0.000442 ± 1.0e-12

After (18d85c7)

1. Same command: 8 passed, 290 deselected

Type

🐛 Bug Fix
✅ Test

Caveats (if any)

Low

• chore(prices): sync Vertex AI prices: 14 models #40955 carries the same map change, so it becomes a no-op on this field once this merges

Final Attestation

• The tests check the right things, including the edge cases, and regressions in the respective real-world customer use-cases are not possible after this PR

Link to Devin session: https://app.devin.ai/sessions/7553267d0ea54bf4925a6df40e7ca640
Open in Devin Desktop: https://app.devin.ai/desktop/session/7553267d0ea54bf4925a6df40e7ca640?variant=devin

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment, CI, and merge conflict monitoring

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[greptile-apps]

Greptile Summary

This PR corrects Gemini 3.5 Flash-Lite priority cached-input pricing for Vertex-backed model entries from $0.05/M to $0.054/M

• Updates both the root pricing map and bundled fallback map
• Covers prefixed and unprefixed Vertex-backed model keys
• Updates the behavioral cost-calculation regression test while leaving Gemini API pricing unchanged

Confidence Score: 5/5

The PR appears safe to merge because the pricing correction is consistently applied and covered by the existing behavioral cost test

Both runtime pricing-map copies update the relevant Vertex-backed entries, and the regression test verifies cached-token billing through the service-tier cost path without changing Gemini API behavior

Important Files Changed

Filename	Overview
model_prices_and_context_window.json	Corrects the priority cache-read rate for both Vertex-backed Gemini 3.5 Flash-Lite model keys
litellm/model_prices_and_context_window_backup.json	Keeps bundled fallback pricing synchronized with the root pricing map
tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py	Updates the Vertex priority-tier behavioral expectation while retaining distinct Gemini API pricing coverage

_{Reviews (1): Last reviewed commit: "fix(cost_map): vertex gemini-3.5-flash-l..." | Re-trigger Greptile}

[codspeed]

Merging this PR will not alter performance

✅ 31 untouched benchmarks

---

_{Comparing litellm_gemini35_flash_lite_vertex_priority_cache_rate (18d85c7) with main (7fd541e)}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[devin-ai-integration]

Closing: main already carries this Vertex value and test expectation, and the direct Gemini row fix lives in rolling PR #41597