OpenAI has released two new open-weight LLMs, gpt-oss-120b and gpt-oss-20b, which are the first open-weight models since GPT-2 in 2019 and can run locally due to clever optimizations. The models' architecture has some interesting details, including the use of Swish activation function, GLU feed forward module, and sliding window attention, and they share similarities with other recent models ...