Disable thinking mode in Jan-v1-4B model
#9
by
vuhaix95
- opened
Currently, the Jan-v1-4B model runs with thinking mode enabled, showing internal reasoning steps. This behavior is inconvenient during testing and real-world deployment. A configuration option should be added to disable thinking mode completely, so that the model only returns the final response.
I believe this is based on the newest Qwen3-4b-thinking model and the latest revision does not have a toggle for thinking anymore.
As such this is an upstream request and - if I were on the Menlo team - I would gently close this as WONTFIX.
This is a thinking model, there is no way to "turn off thinking"