Output from reasoning models not parsed correctly

[vaaale]

LocalAI version: 3.9.0

Environment, CPU architecture, OS, and Version: Not relevant

Describe the bug When using a reasoning model like Qwen-3*, the response is not included in the output the way it would be expected according to the OpenAI api. Instead of placing the content of the .. block in "reasoning_content" it's returned in the normal "content" of the response. This obviously breaks any consumer of the api.

To Reproduce Run qwen3-*b

Expected behavior Localai should return the response according to the OpenAI specification.

Logs N/A

Additional context N/A