A new reasoning model sets state-of-the-art on coding benchmarks

OpenAI released a successor to its reasoning line, with significant improvements on SWE-bench Verified and live codegen evaluations. Early access is rolling out to ChatGPT Pro and the API.