Bug: Chat title generation doesn't work well with thinking models (Qwen3)

[daaain]

Describe the bug

Chat title displays massive thinking token block.

To Reproduce
Steps to reproduce the behavior:

1. Use a Qwen3 model (I have qwen3-30b-a3b running in LM Studio and use it as an OpenAI compatible API)
2. Ask anything
3. Observe the massive thinking block in the title

Expected behavior

The thinking section is hidden.

What would also be great is if it was possible to disable thinking for the title generation, in Qwen3's case I managed to do that for the main answer by adding /no_thinking into .goosehints

Screenshots

Please provide following information:

• OS & Arch: Mac OS 15.4.1 (24E263)
• Interface: [UI/CLI] UI
• Version: [e.g. v1.0.2] 1.0.20
• Extensions enabled: [e.g. Computer Controller, Figma] Developer, Computer Controller
• Provider & Model: [e.g. Google - gemini-1.5-pro] qwen3-30b-a3b running in LM Studio

Additional context
Add any other context about the problem here.

[DOsinga]

fixed in #3710