This researcher turned OpenAI’s open weights model gpt-oss-20b into a non-reasoning ‘base’ model with less alignment, more freedom
1 min read
Summary
OpenAI has released its new open weights AI large language model (LLM) family, gpt-oss, under a permissive Apache 2.0 licence, its first open weights model release since GPT-2 in 2019.
Developers outside the company have already begun to reshape the product, including Jack Morris, a Cornell Tech PhD student, former Google Brain Resident, and current researcher at Meta, who this week unveiled a reworked version of OpenAI’s smaller gpt-oss-20B model, which he claims removes the reasoning behaviour of the model, returning it to a pre-trained base version that offers faster, freer, more uncensored and unconstrained responses.
The new model is available on Hugging Face under a permissive MIT License and can be used for further research or commercial applications.