OpenAI releases GPT-5.3-Codex, merging its coding and general reasoning models

OpenAI/Codex

OpenAI released GPT-5.3-Codex, combining coding ability of GPT-5.2-Codex with reasoning and general knowledge of GPT-5.2 in one model that is 25% faster. The company noted it was its first model that helped build itself: early versions debugged its own training and managed deployment. It set new highs on SWE-Bench Pro and Terminal-Bench coding tests.

• Merges GPT-5.2-Codex coding performance with GPT-5.2 reasoning and general knowledge in single model, 25% faster
• First OpenAI model that helped build itself: early versions debugged own training and managed deployment
• New highs on SWE-Bench Pro and Terminal-Bench coding tests; supports long-duration research and execution tasks
• Handles debugging, deployment, monitoring, writing PRDs, editing copy, user research and computer-using agent tasks beyond code