OpenAI Reduces Codex Model Context Window from 372k to 272k Tokens
Key Takeaways
- ▸Codex context window reduced by 100,000 tokens (27% decrease) from 372k to 272k
- ▸Change included in OpenAI SDK version 0.144 release
- ▸Developers using Codex may need to adjust integration code to accommodate smaller context windows
Summary
OpenAI has reduced the model context size for its Codex code generation model from 372,000 tokens to 272,000 tokens in a recent update. The change was included in the latest version 0.144 release, with updated model metadata merged into the repository on July 18, 2026.
The 100,000-token reduction represents a significant 27% decrease in the maximum context window available for code generation, synthesis, and understanding tasks. This adjustment will impact developers and organizations actively using Codex, requiring them to potentially revise integration code and implementation strategies to accommodate the smaller context window. The change likely reflects OpenAI's optimization efforts to improve inference performance, reduce latency, or better align the model with current production requirements.
- Likely reflects optimization efforts to improve inference performance and latency
