OpenAI released GPT-5.3-Codex for coding, calling it the first model that had been instrumental in creating itself: the Codex team used early versions to debug its own training, manage its own deployment, and diagnose test results and evaluations, and OpenAI wrote that the team "was blown away by how much Codex was able to accelerate its own development." It claimed a new industry high on SWE-Bench Pro and Terminal-Bench and strong results on OSWorld and GDPval. It shipped on paid ChatGPT plans everywhere Codex runs — app, CLI, IDE extension, and web — but not in the API at first. On the 12th a smaller GPT-5.3-Codex-Spark followed as a research preview — text-only, with a 128,000-token context and a claimed 1,000 tokens per second — out of the Cerebras partnership announced on 14 January. Anthropic released Claude Opus 4.6 the same 5 February.