OpenAI Launches GPT-5.3-Codex with 25% Faster Speed and First 'High' Cybersecurity Rating to Challenge Anthropic's Claude
Summary
OpenAI unveils GPT-5.3-Codex with 25% faster speeds and breakthrough 'high' cybersecurity rating, expanding beyond coding to debugging, deployment, and business tasks while launching a Mac desktop app to directly challenge Anthropic's Claude dominance.
Key Points
- OpenAI releases GPT-5.3-Codex to compete directly with Anthropic's Claude Cowork, expanding capabilities beyond code writing to include debugging, deployment, user research, and creating presentations and spreadsheets
- The new model features 25% faster inference, improved coding accuracy on developer benchmarks, interactive agent options, and represents OpenAI's first 'high' cybersecurity-rated model that helped create itself
- OpenAI launches a separate Mac desktop Codex app making the tool more accessible to non-coders, though it remains platform-limited compared to Claude's integrated Windows and Mac availability