This model is pretty good at reasoning
#8
by
espyjump
- opened
I have tried this on the recent rounds of AtCoder ABCs, and it performs quite well! I do wonder if some of it is trained on outputs from gpt oss, as they seem to reason in very similar fashions, as well as giving the code with explanations in the same format, along with a ton of policy checks. It appears to be a dense model, but I suppose it might be good for visual reasoning tasks, although the OCR is not exactly satisfactory, and it seems prone to repeat indefinitely on visual tasks.
Hi @espyjump we distill SFT data from gpt-oss-120b
Please have a look at our technical report - https://arxiv.org/abs/2510.01141