OpenAI has released two new open-weight language models — marking the company’s first open model release since GPT-2 in 2019. These models are optimized for advanced reasoning tasks like coding, math, and health queries — and can even run on