fix(model_prices): add provider-announced deprecation_date to 205 registry entries

[devin-ai-integration]

TLDR

Problem this solves:

• 205 registry entries lack provider-announced deprecation_date
• /model/deprecations (feat(proxy): proactive model deprecation alerts and /model/deprecations endpoint #26900) therefore reports them as healthy
• Operators get no warning for Azure, Anthropic, Vertex, OpenAI retirements

How it solves it:

• Adds deprecation_date from each provider's official retirement doc
• Purely additive: no existing value or field is changed
• Skips ambiguous alias/version mappings rather than guessing

User Flow

Before: a proxy admin who deploys claude-sonnet-4-5, azure/gpt-5, vertex_ai/claude-opus-5 and babbage-002 gets no retirement warning even though all four have published retirement dates

1. They call GET https://litellm-domain/model/deprecations with their master key
2. The response comes back {"deprecated": [], "imminent": [], "upcoming": []}
3. They see nothing to act on, and the Anthropic Sonnet 4.5 deployment silently stops working after its retirement date

After: the same call lists every deployment whose provider has announced a retirement date

1. They call GET https://litellm-domain/model/deprecations with their master key
2. upcoming now contains legacy-babbage (2026-09-28), sonnet-4-5 (2026-09-29), vertex-opus-5 (2027-01-24) and azure-gpt-5 (2027-02-09), each with days_until_deprecation
3. They plan migrations ahead of each date, and Slack deprecation alerts fire for the same deployments

Relevant issues

Supports #26900

Linear ticket

Resolves LIT-5850

What changed, by source

Every date below comes from the provider's own retirement/deprecation doc. Only entries that had no deprecation_date were touched; where a provider doc was ambiguous (unversioned Azure alias covering versions with differing dates, azure_ai/model_router, Gemini-API dates for models exposed under vertex_ai, conflicting rows for gpt-realtime-mini-2025-10-06) the entry was left alone

Group	Entries	Source
Azure OpenAI + Azure AI Foundry (azure/*, azure_ai/*)	96	https://learn.microsoft.com/en-us/azure/foundry/openai/concepts/model-retirement-schedule
OpenAI (openai, text-completion-openai, incl. image-size variants)	45	https://platform.openai.com/docs/deprecations
Vertex AI Gemini / Imagen / Veo / embeddings	24	https://cloud.google.com/vertex-ai/generative-ai/docs/learn/model-versions
Vertex AI Claude partner models	24	https://cloud.google.com/vertex-ai/generative-ai/docs/partner-models/use-claude
Anthropic direct (claude-*)	15	https://docs.claude.com/en/docs/about-claude/model-deprecations
Gemini API (gemini/gemini-robotics-er-1.6-preview)	1	https://ai.google.dev/gemini-api/docs/deprecations

Representative values:

  "claude-sonnet-4-5": {              + "deprecation_date": "2026-09-29"   # Anthropic: not sooner than Sept 29, 2026
  "claude-opus-5": {                  + "deprecation_date": "2027-07-24"   # Anthropic: not sooner than July 24, 2027
  "vertex_ai/claude-opus-5": {        + "deprecation_date": "2027-01-24"   # Vertex: not sooner than January 24, 2027
  "azure/gpt-5": {                    + "deprecation_date": "2027-02-09"   # Azure: gpt-5 v2025-08-07 retires 2027-02-09
  "azure_ai/deepseek-r1": {           + "deprecation_date": "2026-08-13"   # Azure Foundry: DeepSeek-R1 retired 2026-08-13
  "gemini-2.5-pro": {                 + "deprecation_date": "2026-10-20"   # Vertex: retirement October 20, 2026
  "babbage-002": {                    + "deprecation_date": "2026-09-28"   # OpenAI: shutdown 2026-09-28

Note that per-provider dates for the same model family deliberately differ (Anthropic direct claude-opus-5 = 2027-07-24 vs Vertex = 2027-01-24 vs Azure Foundry's own schedule), matching each provider's published date rather than normalising them

Audited with no verifiable change needed: Bedrock (already fully in sync with the AWS model lifecycle table), Groq, Mistral, Cohere, xAI, DeepSeek, Fireworks

Pre-Submission checklist

• I have added meaningful tests
• My PR passes all CI/CD checks (e.g., lint, format, unit tests)
• 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

No new tests: this is a data-only change covered by the existing registry schema tests and the /model/deprecations tests (tests/test_litellm/test_model_prices_schema.py 19 passed, deprecation endpoint + util tests 22 passed, backup-sync tests 5 passed, make pre-commit clean)

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

Shared setup, /tmp/dep_config.yaml:

model_list:
  - model_name: sonnet-4-5
    litellm_params: {model: claude-sonnet-4-5, api_key: fake-key}
  - model_name: azure-gpt-5
    litellm_params: {model: azure/gpt-5, api_key: fake-key, api_base: https://fake.openai.azure.com}
  - model_name: vertex-opus-5
    litellm_params: {model: vertex_ai/claude-opus-5}
  - model_name: legacy-babbage
    litellm_params: {model: babbage-002, api_key: fake-key}
general_settings:
  master_key: sk-1234

Before (4fd7a73)

1. LITELLM_LOCAL_MODEL_COST_MAP=True litellm --config /tmp/dep_config.yaml --port 4001 (run from a worktree at the merge base)
2. curl -s -H 'Authorization: Bearer sk-1234' http://localhost:4001/model/deprecations | python3 -m json.tool

{
    "deprecated": [],
    "imminent": [],
    "upcoming": [],
    "warn_within_days": 30,
    "checked_at": "2026-08-18T13:19:51.395249Z"
}

After (3d523d6)

1. LITELLM_LOCAL_MODEL_COST_MAP=True litellm --config /tmp/dep_config.yaml --port 4000
2. curl -s -H 'Authorization: Bearer sk-1234' http://localhost:4000/model/deprecations | python3 -m json.tool

{
    "deprecated": [],
    "imminent": [],
    "upcoming": [
        {
            "model_name": "legacy-babbage",
            "litellm_model": "babbage-002",
            "deprecation_date": "2026-09-28",
            "days_until_deprecation": 41,
            "status": "upcoming",
            "litellm_provider": "text-completion-openai"
        },
        {
            "model_name": "sonnet-4-5",
            "litellm_model": "claude-sonnet-4-5",
            "deprecation_date": "2026-09-29",
            "days_until_deprecation": 42,
            "status": "upcoming",
            "litellm_provider": "anthropic"
        },
        {
            "model_name": "vertex-opus-5",
            "litellm_model": "vertex_ai/claude-opus-5",
            "deprecation_date": "2027-01-24",
            "days_until_deprecation": 159,
            "status": "upcoming",
            "litellm_provider": "vertex_ai-anthropic_models"
        },
        {
            "model_name": "azure-gpt-5",
            "litellm_model": "azure/gpt-5",
            "deprecation_date": "2027-02-09",
            "days_until_deprecation": 175,
            "status": "upcoming",
            "litellm_provider": "azure"
        }
    ],
    "warn_within_days": 30,
    "checked_at": "2026-08-18T13:17:04.707460Z"
}

Type

🐛 Bug Fix

Caveats (if any)

• Anthropic and Vertex Claude dates are earliest-possible "not sooner than" dates
• Entries whose date already passed now report and alert as deprecated
• Ambiguous unversioned aliases deliberately keep no date

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/77eaacb703734bc1a155b51c5b35f737

---

Note

^{Cursor Bugbot is generating a summary for commit 3d523d6. Configure here.}

• 3d523d6 passes /live-pr-risk

[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

Adds provider-announced deprecation metadata to 205 model registry entries so configured deployments can appear in deprecation reporting and alerts.

• Covers Azure OpenAI and Azure AI Foundry, OpenAI, Vertex AI, Anthropic, and Gemini model entries.
• Keeps the canonical and packaged backup registries synchronized.
• Makes no pricing, routing, authentication, or executable-code changes.

Confidence Score: 5/5

The PR appears safe to merge with no actionable defects identified in the synchronized metadata additions.

The change only adds schema-compatible deprecation dates to matching entries in both registry copies, and no concrete incorrect mapping, malformed value, or consumer incompatibility was established.

Important Files Changed

Filename	Overview
model_prices_and_context_window.json	Adds valid ISO-formatted deprecation dates to 205 canonical registry entries without changing existing fields.
litellm/model_prices_and_context_window_backup.json	Mirrors the canonical registry's 205 deprecation-date additions exactly.

_{Reviews (1): Last reviewed commit: "fix(model_prices): add provider-announce..." | Re-trigger Greptile}

[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 devin/1787058723-registry-deprecation-dates (3d523d6) with litellm_internal_staging (4fd7a73)}

[mateo-berri]

bugbot run

[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 3d523d6. Configure here.}

[mateo-berri]

LGTM