I'm using Groq through the API.
The model outputs usually show reasoning within <think> tokens (if they support reasoning, like qwen3-32b) and this is always part of the completion.
I could not find a way to either:
- set
reasoning_effort="none"(or any other value) in the parameters to change reasoning.
- hide
<think>...</think> tokens.
I'm using Groq through the API.
The model outputs usually show reasoning within
<think>tokens (if they support reasoning, likeqwen3-32b) and this is always part of the completion.I could not find a way to either:
reasoning_effort="none"(or any other value) in the parameters to change reasoning.<think>...</think>tokens.