# speech_model on POST /v2/transcript is deprecated

AssemblyAI API deprecated POST /v2/transcript, body parameter speech_model without giving a shutdown date yet.

AssemblyAI deprecated the body parameter speech_model on POST /v2/transcript on July 10, 2026, without announcing a shutdown date. Requests that supply current model names like universal-2 or universal-3-5-pro in the speech_model field now fail with a 400 error.

## What to use instead

Send speech_models as a list, such as ["universal-3-5-pro", "universal-2"], instead of speech_model to pin a model.

- API: AssemblyAI API (api.assemblyai.com)
- Kind: Parameter removal
- Announced: July 10, 2026
- Takes effect: no date given yet
- Calls after that day: fail
- Calls it hits: `POST /v2/transcript`, `body parameter speech_model`
- Source: [Announcement](https://www.assemblyai.com/changelog)
- Verified: November 6, 2026, against AssemblyAI API's own pages
- Page: https://www.okapi.bot/providers/assemblyai-api/breaking-changes/speech-model-on-post-v2-transcript-is-deprecated/

## AssemblyAI API's words

> Deprecated: Async speech_model Parameter Passing a current model name ( universal-2 , universal-3-5-pro ) to the async speech_model field now returns a 400.

Okapi finds the calls this change hits in your repo and opens the pull request with the fix: https://github.com/apps/okapi-agent