OpenAI’s GPT-5.3-Codex thinks deeper and wider about coding work

OpenAI launched GPT-5.3-Codex, expanding its coding agent to broader work tasks beyond just code. This new model integrates GPT-5.2-Codex's coding prowess with GPT-5.2's reasoning and professional knowledge, operating 25% faster. It can handle complex, long-running tasks involving research, tool use, and intricate planning across general work and software development. OpenAI claims over one million developers use Codex, while Anthropic's Claude Code is also gaining significant traction, with some projecting its use in GitHub commits to reach 20% by 2026. GPT-5.3-Codex achieved top scores on SWE-Bench Pro and Terminal-Bench 2.0, benchmarks for software engineering and terminal skills. Anthropic’s new Claude Opus 4.6 also claims top scores on various industry benchmarks for multidisciplinary reasoning and economically valuable knowledge work. OpenAI's model can process more information and think longer without human intervention, autonomously iterating on game development. Similarly, Claude Opus 4.6 comprehends larger codebases and makes more thoughtful decisions for new code. GPT-5.3-Codex supports the entire software lifecycle, including debugging, deploying, and monitoring, and extends to tasks like creating slide decks and analyzing data. It matches GPT-5.2 on knowledge-work evaluations and showed improved accuracy on computer use tests. This model is also designated "high capability" for cybersecurity and can identify software vulnerabilities, with OpenAI committing $10 million in API credits for cyber defense. GPT-5.3-Codex is currently available to paid ChatGPT subscribers through various interfaces, with API access planned for the future.