OpenAI recommends GPT-5.3-Codex as the default agentic coding model
OpenAI now recommends GPT-5.3-Codex as the default Codex model, signaling a step-up in agentic coding and reasoning for real-world engineering. The official Codex Models page highlights GPT-5.3-Codex as the most capable, with GPT-5.2-Codex as predecessor and a smaller GPT-5.1-Codex-mini option for cost-sensitive tasks [OpenAI Codex Models](https://developers.openai.com/codex/models/)[^1]. An anecdotal report describes spending $10,000 to automate research with Codex, indicating emerging large-scale usage patterns [Practitioner report](https://links.tldrnewsletter.com/J7poJAf substantial Codex-driven automation and spend.