fix(pricing): add undated azure gpt-audio-mini and gpt-realtime-mini aliases

[devin-ai-integration]

Relevant issues

Fixes #33170

Internal replacement branch for OSS PR #33291 by Mihidum Hettiyahandi (@mihidumh); original authorship is preserved on the commits

Linear ticket

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

Screenshots / Proof of Fix

Captured live against a proxy on localhost:4000 at commit f3464b15d5, started with LITELLM_LOCAL_MODEL_COST_MAP=True so the resolution comes from this branch's bundled cost map

Config declares the two undated deployments with dummy Azure creds (no request is needed, the lookup is metadata-only):

model_list:
  - model_name: azure-gpt-audio-mini
    litellm_params: {model: azure/gpt-audio-mini, api_key: dummy, api_base: https://dummy.openai.azure.com}
  - model_name: azure-gpt-realtime-mini
    litellm_params: {model: azure/gpt-realtime-mini, api_key: dummy, api_base: https://dummy.openai.azure.com}

/v1/model/info now resolves pricing and limits for the undated aliases:

$ curl -s http://localhost:4000/v1/model/info -H "Authorization: Bearer sk-1234" | jq ...
azure/gpt-audio-mini    input=6e-07 output=2.4e-06 input_audio=1e-05 max_in=128000 max_out=16384 mode=chat
azure/gpt-realtime-mini input=6e-07 output=2.4e-06 input_audio=1e-05 max_in=32000  max_out=4096  mode=chat

Before the fix these undated names were absent from the cost map, so the same lookup returned zeroed/None pricing and no token limits

Type

🐛 Bug Fix

Changes

Adds undated azure/gpt-audio-mini and azure/gpt-realtime-mini entries to model_prices_and_context_window.json, mirroring the existing dated -2025-10-06 entries (costs, token limits, supported modalities, endpoint metadata). litellm/model_prices_and_context_window_backup.json is synced via ci_cd/check_files_match.py so the two files match

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/2a3db9e7f3234370988a77b458d0a202

[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

[greptile-apps]

Greptile Summary

This PR adds undated azure/gpt-audio-mini and azure/gpt-realtime-mini entries to both cost map files, mirroring the existing -2025-10-06 dated entries exactly, so that Azure deployments using the versionless model name receive correct pricing and token limits instead of zeroed values.

• model_prices_and_context_window.json / backup: Two new entries added as exact copies of their dated counterparts; all cost fields, token limits, and capability flags are identical.
• Backup sync: The backup file is additionally brought in sync with several other models already present in the primary file (gemini-3-pro-image, zai/glm-*, etc.) as expected by ci_cd/check_files_match.py.
• Test: test_azure_audio_price_aliases.py forces local cost-map loading and asserts all four audio cost fields are non-zero and match the dated entry, directly covering the original bug symptom.

Confidence Score: 5/5

Safe to merge — the change is purely additive data in JSON cost maps with no logic changes.

Both new entries are verified field-for-field copies of their existing dated counterparts, the backup file is kept in sync by the CI check, and the new regression test directly exercises the reported bug symptom using only local data.

No files require special attention.

Important Files Changed

Filename	Overview
model_prices_and_context_window.json	Adds two new undated alias entries (azure/gpt-audio-mini and azure/gpt-realtime-mini) that are exact field-for-field copies of their dated -2025-10-06 counterparts; no values differ.
litellm/model_prices_and_context_window_backup.json	Synced to match the primary cost map; includes the two new Azure aliases plus several other model entries (gemini-3-pro-image, gemini-3.1-flash-image, zai/glm-*, darkbloom/*) that were already in the main file but missing from the backup.
tests/test_litellm/test_azure_audio_price_aliases.py	New regression test that loads the local cost map via LITELLM_LOCAL_MODEL_COST_MAP=True, then asserts all four audio cost fields on each undated alias are non-zero and match their dated counterpart.

_{Reviews (1): Last reviewed commit: "chore: sync model_prices_and_context_win..." | 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]

Tested live against a local proxy on localhost:4000 started with LITELLM_LOCAL_MODEL_COST_MAP=True so pricing resolves from this branch's cost map, then queried /v1/model/info

curl -s http://localhost:4000/v1/model/info -H "Authorization: Bearer sk-1234"

azure/gpt-audio-mini
  input_cost_per_token   = 6e-07
  output_cost_per_token  = 2.4e-06
  input_cost_per_audio   = 1e-05
  max_input_tokens       = 128000
  max_output_tokens      = 16384
  mode                   = chat
azure/gpt-realtime-mini
  input_cost_per_token   = 6e-07
  output_cost_per_token  = 2.4e-06
  input_cost_per_audio   = 1e-05
  max_input_tokens       = 32000
  max_output_tokens      = 4096
  mode                   = chat

Both undated aliases now resolve full pricing and token limits; before this change the undated names were absent from the map and returned 0/None

Full walkthrough and test report: https://app.devin.ai/sessions/2a3db9e7f3234370988a77b458d0a202