chore(model_prices): add Gemini Veo, Mistral and Azure Claude 4.5 deprecation dates

[devin-ai-integration]

TLDR

Problem this solves:

• Six Gemini Veo and Mistral models have announced shutdowns but no deprecation_date
• Three azure_ai Claude 4.5 rows carry a stale Azure retirement date

How it solves it:

• Add deprecation_date to the six direct-provider keys from the provider deprecation tables
• Move the azure_ai Claude 4.5 dates to the current Azure schedule (absorbed from chore(cost-map): take azure_ai claude 4.5 retirement dates from the Azure schedule #43846)

This is the rolling registry PR for the 2026-09-30 audit run. Every change is registry data only, no code path changed

User Flow

Before: a proxy admin running these models gets no warning from the deprecations report before the provider shuts them off

1. They configure gemini/veo-3.1-generate-preview, mistral/mistral-ocr-4-0 and azure_ai/claude-sonnet-4-5 on the proxy
2. They send GET https://litellm-domain/model/deprecations with the master key
3. The Veo and Mistral models are missing from every bucket, and azure_ai/claude-sonnet-4-5 is reported as retiring on 2026-10-19

After: the same report lists the right shutdown dates

1. Same config and the same GET https://litellm-domain/model/deprecations
2. gemini/veo-3.1-generate-preview shows 2026-10-22 and mistral/mistral-ocr-4-0 shows 2026-09-30
3. azure_ai/claude-sonnet-4-5 now shows 2026-11-15, the date on the Azure schedule

Sources

All values below were read from the raw HTML table rows on 2026-09-30, with r.jina.ai as a second reader. Both agreed on every row

Gemini API: https://ai.google.dev/gemini-api/docs/deprecations, columns Model, Release date, Shutdown date, Replacement

Key	deprecation_date	Raw row
gemini/veo-3.1-generate-preview	2026-10-22	veo-3.1-generate-preview, October 15, 2025, October 22, 2026, gemini-omni-1.1-flash
gemini/veo-3.1-fast-generate-preview	2026-10-22	veo-3.1-fast-generate-preview, October 15, 2025, October 22, 2026, gemini-omni-1.1-flash
gemini/veo-3.1-lite-generate-preview	2026-10-22	veo-3.1-lite-generate-preview, March 31, 2026, October 22, 2026, gemini-omni-1.1-flash

Mistral: https://docs.mistral.ai/getting-started/models/models_overview, deprecated models table, the Deprecation and Retirement column holds two dates and the second is the retirement date. The model cards (https://docs.mistral.ai/models/zai-glm-5-2, https://docs.mistral.ai/models/leanstral-1-5, https://docs.mistral.ai/models/ocr-4-0) mark each model deprecated as of 9/29/2026. deprecation_date holds the retirement date, the day calls stop working, matching how OpenAI and Gemini shutdown dates are recorded

Key	deprecation_date	Raw row
mistral/zai-glm-5-2	2026-10-31	Z.ai GLM 5.2, zai-glm-5-2, 9/29/2026, 10/31/2026, Z.ai GLM 5.3
mistral/labs-leanstral-1-5	2026-09-30	Leanstral 1.5, labs-leanstral-1-5, 9/29/2026, 9/30/2026
mistral/mistral-ocr-4-0	2026-09-30	OCR 4.0, mistral-ocr-4-0, 9/29/2026, 9/30/2026, OCR 4.1

Azure: https://learn.microsoft.com/en-us/azure/foundry/openai/concepts/model-retirement-schedule, Anthropic rows

Key	Old	New	Raw row
azure_ai/claude-sonnet-4-5	2026-10-19	2026-11-15	claude-sonnet-4-5, 1, GA, 2026-11-15
azure_ai/claude-opus-4-5	2026-10-19	2026-11-24	claude-opus-4-5, 1, GA, 2026-11-24
azure_ai/claude-haiku-4-5	2026-10-19	2026-11-15	claude-haiku-4-5, versions 1 and 2, GA, 2026-11-15

These three rows also gain a source field pointing at the schedule, as #43846 did

Other open registry PRs reviewed this run

Absorbed: #43846 (Azure Claude 4.5 retirement dates), every date re-checked against the raw Azure schedule rows above

Left open, not absorbed:

• chore(cost-map): add fireworks priority prices for ember-1, nemotron and glm 5.3 us rows #43811 (Fireworks priority prices): the raw https://docs.fireworks.ai/serverless/pricing fetch has no per-model rows for Ember-1, Nemotron or GLM 5.3 (US), so the priority prices could not be confirmed
• feat(model_prices): add deepseek-v3 to model price maps (#43660) #43753 (DeepSeek V3): https://api-docs.deepseek.com/quick_start/pricing lists only deepseek-flash and deepseek-v4-pro, no V3 price row to verify against
• feat(models): sync together_ai model registry #41775 (Together AI sync), fix(prices): fix xai/grok-4 capabilities #41376 (xAI Grok flags), fix: correct cache pricing for Claude 3 Haiku CRIS variants on Bedrock #40368 (Bedrock Claude cache pricing): same reasons as the 2026-09-26 run, nothing on the provider pages changed since
• fix(cost_map): mark fireworks_ai/minimax-m3 as vision and pin capabilities verified against live calls #43390 (Fireworks capability pins), feat: add CompactifAI model pricing and capabilities #43457 (CompactifAI), docs: add Phoenix Grove (pgsgrove) model pricing entries #40755 (PGS Grove): code or provider changes that also touch the JSON

Checked and left unchanged: OpenAI GPT-6.1 Sol, GPT-6 Astra (all tiers including ultrafast and long context) and the deprecations page match the map. Anthropic Claude Sonnet 5.5 pricing matches, and its "Not sooner than" date on the deprecations page is not a retirement announcement so it was not recorded. Vertex Veo 3.1 GA dates match. Mistral dates were applied only to mistral/ keys, the Azure and Vertex copies of those models follow their own platform schedules. mistral/glm-5-2 and mistral/labs-leanstral-1-5-1 are not named in the Mistral table and stay undated. Cohere embed-v5.0-fast and embed-v5.0-pro are new on the Cohere models page but have no raw price row, so they were not added. Bedrock pricing rows for new models render as template placeholders in raw HTML and were not used. ci_cd/cost_map_pins.json is not on main, and no capability flag was changed

Pre-Submission checklist

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

• I have added meaningful tests (data only change, covered by the existing schema and deprecation tests)
• The handful of test files covering my change pass locally (tests/unit/test_model_prices_schema.py, tests/test_litellm/proxy/common_utils/test_model_deprecation.py, tests/test_litellm/proxy/test_model_deprecations_endpoint.py, ci_cd/check_files_match.py, ci_cd/cost_map_guard.py, schema generator --check with no diff)
• 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 (registry data for one audit run)
• I have received a Greptile Confidence Score of at least 4/5 before requesting a maintainer review

Screenshots / Proof of Fix

Source check, raw Gemini deprecations table:

$ curl -sL https://ai.google.dev/gemini-api/docs/deprecations -o dep.html   # then the <tr>/<td> row parser
['veo-3.1-lite-generate-preview', 'March 31, 2026', 'October 22, 2026', 'gemini-omni-1.1-flash']
['veo-3.1-generate-preview', 'October 15, 2025', 'October 22, 2026', 'gemini-omni-1.1-flash']
['veo-3.1-fast-generate-preview', 'October 15, 2025', 'October 22, 2026', 'gemini-omni-1.1-flash']

Source check, raw Mistral models table (the date cell is two spans, deprecation then retirement):

$ curl -sL https://docs.mistral.ai/getting-started/models/models_overview -o mistral.html   # same parser
['Z.ai GLM 5.2 ↗', '5.2', 'zai-glm-5-2', '9/29/2026 10/31/2026', 'Z.ai GLM 5.3']
['Leanstral 1.5 ↗', '1.5', 'labs-leanstral-1-5', '9/29/2026 9/30/2026', '']
['OCR 4.0 ↗', '4.0', 'mistral-ocr-4-0', '9/29/2026 9/30/2026', 'OCR 4.1']

Source check, raw Azure retirement schedule:

$ curl -sL https://learn.microsoft.com/en-us/azure/foundry/openai/concepts/model-retirement-schedule -o az.html   # same parser
['claude-haiku-4-5', '1', 'GA', '2026-11-15']
['claude-haiku-4-5', '2', 'GA', '2026-11-15']
['claude-opus-4-5', '1', 'GA', '2026-11-24']
['claude-sonnet-4-5', '1', 'GA', '2026-11-15']

Type

Refactoring

Caveats (if any)

Low

• mistral/labs-leanstral-1-5 and mistral/mistral-ocr-4-0 retire today, so they land straight in the deprecated bucket
• A pre-existing tests/unit/test_utils.py static schema failure on gpt-6-astra ultrafast fields also fails on main

Link to Devin session: https://app.devin.ai/sessions/df8f278450574b27873795ce256c627e
Open in Devin Desktop: https://app.devin.ai/desktop/session/df8f278450574b27873795ce256c627e?variant=devin

---

Note

Low Risk
Metadata-only JSON updates to deprecation dates and documentation sources; no application code paths change.

Overview
Updates the model price registry so GET /model/deprecations and related tooling reflect current provider retirement schedules—no runtime routing or pricing logic changes.

Azure AI Claude 4.5 (claude-haiku-4-5, claude-sonnet-4-5, claude-opus-4-5): deprecation_date moves from 2026-10-19 to 2026-11-15 (haiku/sonnet) and 2026-11-24 (opus), with a new source URL on the Azure model retirement schedule page.

Gemini Veo 3.1 preview video models (veo-3.1-generate-preview, fast, lite): adds deprecation_date: 2026-10-22.

Mistral (zai-glm-5-2, mistral-ocr-4-0, labs-leanstral-1-5): adds retirement deprecation_date values (2026-10-31 and 2026-09-30 per model).

The same edits are applied to model_prices_and_context_window.json and litellm/model_prices_and_context_window_backup.json so the pair stays in sync.

^{Reviewed by Cursor Bugbot for commit d6b9dd5. Bugbot is set up for automated code reviews on this repo. Configure here.}

[devin-ai-integration]

I'll fix CI failures and address comments from users with write access. I'll skip comments containing "(aside)".

• Disable automatic comment, CI, and merge conflict monitoring

[mateo-berri]

@greptileai

[mateo-berri]

bugbot run

[greptile-apps]

Confidence Score: 5/5

[Medium risk] Updates model deprecation dates and metadata in pricing tables.

The PR appears safe to merge based on the changed registry entries and their existing consumers.

Summary

The PR adds retirement dates for three Gemini Veo and three Mistral models, and updates dates and source links for three Azure AI Claude 4.5 models.

_{Reviews (1) · Last reviewed commit: "chore(model_prices): take azure_ai Claud..."}

[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.}

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit d6b9dd5. Configure here.}

[codspeed]

Merging this PR will not alter performance

✅ 31 untouched benchmarks

---

_{Comparing litellm_registry_audit_20260930 (d6b9dd5) with main (6bc17f9)¹}

Footnotes

1. No successful run was found on main (b370996) during the generation of this report, so 6bc17f9 was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!