fix(vertex_ai): add Gemini 3.7 Flash regions

[pmerlin1]

TLDR

Problem this solves:

• Gemini 3.7 Flash falls back to unsupported us-central1 when no location is supplied
• Saved deployments return Vertex AI publisher-model 404 errors

How it solves it:

• Registers Gemini 3.7 Flash's supported Vertex AI locations: global, us, and eu
• Defaults a missing location to global
• Preserves explicitly configured locations instead of silently changing their residency boundary
• Tests missing, supported, and unsupported explicit locations

User Flow

Before: a proxy admin saves Gemini 3.7 Flash, but a deployment without a persisted location routes to unsupported us-central1

1. They open /ui/?page=models, add vertex_ai/gemini-3.7-flash, enter their Vertex project, set the location to global, and see the connection test succeed
2. They save the deployment and send POST /v1/chat/completions with "model": "gemini-3.7-flash"
3. The request returns 404 because the publisher-model path contains locations/us-central1

After: a missing saved location defaults to a supported endpoint without overriding an explicit regional setting

1. They save or configure vertex_ai/gemini-3.7-flash
2. A deployment with no location uses global
3. Explicit global, us, eu, or regional values remain unchanged

Relevant issues

Fixes #37989

Linear ticket

Pre-Submission checklist

Please complete all items before asking a LiteLLM maintainer to review your PR

• 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. Leave the suites (make test-unit-*, make test-unit) to CI: it finishes in ~15 minutes where a laptop takes an hour or more
• My PR passes all required CI/CD checks (e.g., lint, schema.d.ts sync check, etc.)
• 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 (Greptile reviews automatically once the PR is opened; only comment @greptileai to re-request a review after pushing changes)

Delays in PR merge?

If you're seeing a delay in your PR being merged, ping the LiteLLM Team on Slack (#pr-review)

Screenshots / Proof of Fix

Before (75613bf)

1. Run git show 75613bf22f:model_prices_and_context_window.json | jq '."vertex_ai/gemini-3.7-flash".supported_regions // null'
2. The command prints null, so LiteLLM has no model-specific location metadata
3. A live saved-model request returns 404 with locations/us-central1 in the Vertex AI publisher-model path, as reproduced in [Bug]: Vertex AI Gemini 3.7 Flash defaults to unsupported us-central1 location #37989

After (e05d99d)

1. Run git show e05d99d345:model_prices_and_context_window.json | jq '."vertex_ai/gemini-3.7-flash".supported_regions // null'
2. The command prints ["global", "us", "eu"], matching the Gemini 3.7 Flash model page
3. Missing locations resolve to global; explicit locations remain unchanged

Type

🐛 Bug Fix
✅ Test

Caveats (if any)

• Live after-proof requires deploying a build from this branch
• This PR does not change model-save parameter persistence
• Explicit unsupported locations are preserved and can still produce a provider error; LiteLLM no longer silently broadens them

QA runbook

• tests/test_litellm/llms/vertex_ai/test_vertex_ai_common_utils.py::test_get_vertex_region_gemini_3_7_flash verifies Gemini 3.7 Flash location selection
  • Run the parametrized test with no location and expect global
  • Run it with global, us, and eu and expect each unchanged
  • Run it with us-central1 and expect us-central1 unchanged
• Qwen and Gemma resolver tests verify the same explicit-location contract for global-only model metadata

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

[veria-ai]

PR overview

All previously flagged issues have been addressed. No open security concerns remain on this pull request.

Security review

No open security issues remain on this pull request.

Fixed/addressed: 1 · PR risk: 0/10

[greptile-apps]

Greptile Summary

This follow-up synchronizes Gemini 3.7 Flash region metadata into both model catalogs while preserving explicitly configured Vertex AI locations.

• Registers global, us, and eu as supported regions for Gemini 3.7 Flash.
• Uses global when this model has no configured location.
• Preserves explicit locations and updates the corresponding Gemini, Gemma, and Qwen tests.

Confidence Score: 5/5

The PR appears safe to merge.

No blocking failure remains.

Important Files Changed

Filename	Overview
litellm/llms/vertex_ai/vertex_llm_base.py	Region resolution now uses catalog metadata only for missing locations and preserves every explicit location.
model_prices_and_context_window.json	Adds Gemini 3.7 Flash region metadata to the root model registry.
litellm/model_prices_and_context_window_backup.json	Synchronizes the packaged runtime catalog with the root registry.
tests/test_litellm/llms/vertex_ai/test_vertex_ai_common_utils.py	Covers missing, supported, and explicitly configured region resolution for Gemini 3.7 Flash.
tests/test_litellm/llms/vertex_ai/vertex_ai_partner_models/gemma/test_vertex_ai_gemma_global_endpoint.py	Updates Gemma coverage to enforce preservation of explicit regions.
tests/test_litellm/llms/vertex_ai/vertex_ai_partner_models/qwen/test_vertex_ai_qwen_global_endpoint.py	Updates Qwen coverage to enforce preservation of explicit regions.

_{Reviews (3): Last reviewed commit: "fix(model_cost): sync Gemini 3.7 regions..." | Re-trigger Greptile}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[devin-ai-integration]

Reviewed while consolidating registry changes into the rolling PR #37902. Not rolling this one in yet, reason in that PR description under 'Source PR disposition for this run'. Leaving it open for the author to respond.

[pmerlin1]

Updated the resolver and tests to preserve explicit regions, and confirmed Google lists only global, us, and eu. @devin-ai-integration @greptileai please review again

[pmerlin1]

Synced the packaged local model catalog with the root registry, fixing the CI failure while preserving explicit regions. @devin-ai-integration @greptileai please review again

[codspeed]

Merging this PR will not alter performance

✅ 31 untouched benchmarks

---

_{Comparing pmerlin1:codex/gemini-37-vertex-regions (cca5fea) with litellm_internal_staging (ca0b951)}