Archive — US — techcrunch
OpenAI upgrades Codex with a new version of GPT-5
OpenAI has launched GPT-5-Codex, an upgraded version of its AI coding agent Codex, enhancing its 'thinking' capabilities for improved performance in coding tasks. Now available to various ChatGPT users, this version aims to compete in the crowded AI coding tool market, outperforming prior benchmarks and offering effective code reviews, reflecting OpenAI's innovative developments in AI technology.
Key points
- OpenAI releases GPT-5-Codex for its Codex AI coding agent.
- The new model can dynamically adjust how much time it spends on coding tasks.
- GPT-5-Codex shows improved performance on agentic coding benchmarks such as SWE-bench Verified.
- It surpasses GPT-5 in coding tasks and refactoring benchmarks.
- Users can access the update via terminals, IDEs, GitHub, and ChatGPT.
- Feedback from software engineers highlights fewer incorrect review comments and more impactful suggestions.
Why it matters
- The launching of GPT-5-Codex aligns with growing demand for AI coding tools.
- The market is now competitive with products like Claude Code and GitHub Copilot.
- In 2025, Cursor achieved significant financial success, indicating a vigorous industry.
- OpenAI's emphasis on real-time decision-making in AI reflects current trends in AI adaptability and performance.
The release of GPT-5-Codex marks a significant advancement in AI coding capabilities, poised to enhance productivity while addressing competitive pressures within the rapidly evolving AI coding landscape.