GPT-5-Codex Model Update: Better Edits & Efficiency
GPT-5-Codex gets more reliable file edits, better collaboration with your changes, and fewer destructive git operations. Plus 3% efficiency gains.
TL;DR
- apply_patch is now more reliable for file edits
- Fewer destructive git operations like reset
- Better collaboration when you edit files alongside the model
- 3% faster and more efficient overall
New
- Improved apply_patch reliability — File edits are now more stable and predictable, reducing failed patches and rollbacks.
- Collaborative file handling — The model better respects your edits and avoids clobbering changes you've made while it's working.
Fixed
- Fewer destructive git actions — Reduced aggressive operations like
git resetthat could lose work. - Performance gains — 3% improvement in execution time and token usage across the board.
Update via your Codex client or CLI. See the full changelog for details.
Source: Codex
