Models with high and/or controllable reasoning effort

Models
1

I have used Open AI’s gpt-oss-20b Model. In LMStudio, it is very easy to change the reasoning effort between Low, Medium and High, it is right there in the GUI. With reasoning effort set to I high, I think it delivers very good results and is still fast because it fits into 16Gbyte VRAM. I thing, I have used larger models that take longer, with less satisfactory results. I wish, I could just tell other models to think for shorter or longer periods of time, depending on the task.
Are there other models that support this feature?

I use AI for creative writing, mostly, at the moment. Last one I have tried is Qwq 32B Q6_K and I think it makes stupid mistakes in logic and continuity in creative writing.

1 Like
2

LM Studio’s reasoning effort feature appears to simply wrap the system prompt passed to the model, so in that sense, there are several other models capable of similar switching.

3

Wow, thanks a lot! I will need some time to understand all this, as I am quite new to the entire thing. But I really appreciate the effort!

1 Like