feat(azure-ai): add latest catalog model metadata

[emerzon]

TLDR

Problem this solves:

• Eight exact Azure AI catalog IDs lack model metadata
• Missing prices and limits weaken routing and spend tracking

How it solves it:

• Adds modes, prices, limits, capabilities, and retirement dates
• Keeps primary and bundled catalogs covered by focused tests

User Flow

Before: a developer cannot inspect complete built-in metadata for a newly configured Azure AI catalog deployment

1. They configure azure_ai/FW-GLM-5.3 for an Azure AI deployment
2. They call GET https://litellm-domain/v1/model/info
3. The deployment has no built-in price, context, retirement, or capability metadata

After: the same developer can inspect the model's published metadata before sending traffic

1. They configure azure_ai/FW-GLM-5.3 for an Azure AI deployment
2. They call GET https://litellm-domain/v1/model/info
3. The deployment includes its price, context, retirement, and capability metadata

Relevant issues

Linear ticket

Sources

• Catalog metadata: DeepSeek, GLM, MAI Thinking, Codestral, Kimi, Grok, MAI Image, and Mistral OCR
• Azure pricing: Microsoft, Mistral, and Kimi
• Provider pricing: Fireworks DeepSeek, Fireworks GLM, and xAI Grok
• Lifecycle dates: Microsoft model retirement schedule

Validation notes

• Native Azure Kimi-K3 deployments provisioned successfully in two regions
• Both regional native inference endpoints returned HTTP 404
• The live gateway removed those routes and retained working FW-Kimi-K3
• The Kimi-K3 entry remains published catalog metadata, not live-route validation

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

This PR adds catalog metadata only; live inference varies by model and Azure deployment

Type

New Feature
Test

Caveats (if any)

Medium

• Fireworks DeepSeek pricing is provisional until Azure publishes its meter
• Fireworks GLM pricing is provisional until Azure publishes its meter
• xAI Grok pricing is provisional until Azure publishes its meter
• Native Kimi-K3 metadata does not imply working Azure inference
  • Two regional deployments provisioned successfully
  • Both regional inference endpoints returned HTTP 404
  • Live routes were removed while FW-Kimi-K3 remained active

Low

• MAI Image uses the live Azure catalog retirement date
• Microsoft's public schedule currently reports 2026-10-01 instead

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

[devin-ai-integration]

PR #39422 labeled risk:medium (3 files, +570/-0: additive model pricing metadata + tests; exceeds low-risk size threshold, no sensitive area). No enterprise label, so Linear routing did not run.

[codspeed-hq]

Merging this PR will not alter performance

✅ 31 untouched benchmarks

---

_{Comparing emerzon:litellm_azure_ai_catalog_models (1a8c889) with litellm_internal_staging (6c5fb0e)}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[devin-ai-integration]

#39388 now carries Codestral-2501, MAI-Thinking-1 and mistral-ocr-4-0 from here. The other five entries have no Azure Retail Prices meter yet, so they stay unabsorbed

[devin-ai-integration]

Rolling registry PR #39388 now carries Codestral-2501, MAI-Thinking-1 and mistral-ocr-4-0 with Azure-published fields; the other five rows still need an Azure meter

[mateo-berri]

Closing: the verifiable rows landed in #39388; the remaining entries have no Azure meter or pricing row to check against. Reopen with an official source.