feat(pricing): add gemini-live-2.5-flash-native-audio to model cost map

[mubashir1osmani]

TLDR

Problem this solves:

• vertex_ai/gemini-live-2.5-flash-native-audio logged zero usage and $0 spend
• The model was missing from the cost map entirely
• Missing entry also meant TEXT modality was never coerced to AUDIO, so sessions hung

How it solves it:

• Adds the vertex_ai and gemini entries with GA pricing and gemini_native_audio: true
• Adds chirp-3 and veo-3.1-lite-generate-001 for vertex_ai in the same pass
• Native-audio detection now resolves the model, so responseModalities is coerced to AUDIO
• Carries the same caching and search-grounding costs as the preview alias of the same model

User Flow

Before: a proxy operator using vertex_ai/gemini-live-2.5-flash-native-audio sees no usage logged

1. Client connects to wss://litellm-domain/v1/realtime?model=vertex_ai/gemini-live-2.5-flash-native-audio
2. Client sends session.update with modalities: ["text"]
3. The session never completes a turn, so no response.done arrives
4. https://litellm-domain/ui/?page=logs shows the request at $0 spend with no token counts

After: the same connection produces a completed response with real usage

1. Client connects to wss://litellm-domain/v1/realtime?model=vertex_ai/gemini-live-2.5-flash-native-audio
2. Client sends session.update with modalities: ["text"]
3. response.done arrives with non-zero audio token counts
4. https://litellm-domain/ui/?page=logs shows the request at non-zero spend, billed at $3.00/M audio in and $12.00/M audio out

Relevant issues

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. 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)

Screenshots / Proof of Fix

Proof requires a live proxy with this model deployed; connecting once it is available on the gateway

Type

🐛 Bug Fix

Caveats (if any)

Low

• Gemini-provider tpm/rpm copied from the dated preview alias of the same model

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

[greptile-apps]

Greptile Summary

Adds Gemini Live 2.5 Flash native-audio catalog metadata so realtime model detection and audio-token spend calculation recognize the GA model.

• Adds the Gemini-prefixed model entry with native-audio capability and GA pricing metadata.
• Advertises /v1/realtime support on the Vertex AI catalog entry.
• Extends realtime transformation tests to cover bare, Gemini-prefixed, and Vertex-prefixed model names.

Confidence Score: 5/5

The PR appears safe to merge.

No blocking failure remains.

Important Files Changed

Filename	Overview
model_prices_and_context_window.json	Adds the Gemini API catalog entry and realtime endpoint metadata for the GA native-audio model; no blocking issue identified.
litellm/model_prices_and_context_window_backup.json	Keeps the packaged fallback catalog synchronized with the primary model-cost map.
tests/test_litellm/llms/gemini/realtime/test_gemini_realtime_transformation.py	Adds catalog-presence and audio-only detection coverage for the new model spellings.

_{Reviews (2): Last reviewed commit: "fix(pricing): add tpm/rpm to gemini/gemi..." | Re-trigger Greptile}

[mubashir1osmani]

bugbot run

[mubashir1osmani]

@greptile-apps

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

Want higher recall? High effort reviews run extra passes and find more bugs. A team admin can switch effort levels in the Cursor dashboard.

^{Reviewed by Cursor Bugbot for commit 8fa9622. Configure here.}

[cursor]

Gemini catalog entry missing sibling cost fields

Medium Severity

The new gemini/gemini-live-2.5-flash-native-audio entry sets supports_web_search to true but omits search_context_cost_per_query, and also drops supports_prompt_caching. Both the Vertex GA twin and the Gemini preview alias include those fields, so search grounding on this alias is billed at $0 and caching is treated as unsupported.

Additional Locations (1)

• litellm/model_prices_and_context_window_backup.json#L20423-L20461

 

^{Reviewed by Cursor Bugbot for commit 8fa9622. Configure here.}

[mubashir1osmani]

Good catch, fixed in d517397. Added cache_read_input_token_cost, supports_prompt_caching, and search_context_cost_per_query, matching the preview alias and both Vertex twins.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[codspeed-hq]

Merging this PR will not alter performance

✅ 31 untouched benchmarks

---

_{Comparing mubashir1osmani:litellm_gemini_live_native_audio_pricing (d517397) with litellm_internal_staging (67c7b97)¹}

Footnotes

1. No successful run was found on litellm_internal_staging (7083c47) during the generation of this report, so 67c7b97 was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩

[devin-ai-integration]

Superseded by the rolling registry PR #38693, which absorbs this change (core pricing verified against https://ai.google.dev/gemini-api/docs/pricing; cache_read/tpm/rpm dropped as not published there). Standing policy: one open model-pricing registry PR at a time.