[FEAT]: Add a "Thinking Mode" Toggle for Hybrid Reasoning Models (DeepSeek, Qwen, Gemini) #4594
Description
What would you like to see?
Dear AnyThingLLM Development Team,
First and foremost, I would like to express my sincere gratitude for developing such an outstanding project. AnyThingLLM provides us with a powerful and flexible multi-model management experience.
I am writing to suggest a feature enhancement. With the widespread adoption of hybrid reasoning models like DeepSeek v3.1/v3.2, qwen3 next, and Gemini 2.5 Flash, their "deep thinking mode" can indeed deliver more accurate and in-depth responses in certain scenarios. However, for daily use or situations requiring faster response times, users might prefer a quicker response mode.
In Cherry Studio, we have already seen an elegant implementation of a similar feature – a simple toggle switch that allows users to flexibly switch between thinking mode and standard mode based on actual needs. This design is both intuitive and practical.
Therefore, I sincerely recommend implementing a similar toggle switch feature in AnyThingLLM, enabling users to:
Enable thinking mode when deep analysis is required
Switch to standard mode when pursuing faster response times
Flexibly adjust according to different usage scenarios
Such a feature would significantly enhance AnyThingLLM's practicality and user experience, making model usage more convenient and effective.
Once again, thank you for your hard work and exceptional contributions. AnyThingLLM has become an indispensable tool, and I sincerely wish the project continued success and growth, bringing value to more users!
Best regards,
A dedicated AnyThingLLM user
