Has anyone managed to get it working in high mode?
#68
by
haejunlee
- opened
It seemd the model does not work in high mode even with system prompt "Reasoning: high"
I tried to add "Reasoning: high" to the system prompt (and developer message too) and even changed the default reasoning mode to high in the chat template, but none of those actually worked.
Reasoning is still very short and the benchmark scores are much lower than the model card.
Even Artifical Analysis scores are very different from the model card.
https://artificialanalysis.ai/models/gpt-oss-120b
Has anyone managed to get it working in high mode?
It seems I should add "reasoning_effort" parameter when calling a client code.
haejunlee
changed discussion status to
closed