openai codex developer-tools code-generation

OpenAI reduces Codex Model Context Size from 372k to 272k

Backport refreshed bundled model metadata to 0.144 by sayan-oai · Pull Request #33972 · openai/codex

github.com

July 19, 2026

1 min read

🔥🔥🔥🔥🔥

64/100

Summary

Backporting refreshed bundled model metadata to version 0.144 is proposed in a pull request. Suggestions cannot be applied while the pull request is closed, and multiple suggestions cannot be batched if they involve deleted lines or subsets of changes.

Key Takeaways

Suggestions cannot be applied while the pull request is closed.
Only one suggestion per line can be applied in a batch.
Applying suggestions on deleted lines is not supported.
Outdated suggestions cannot be applied.

Read original article

View story →

Coverage & discussions

Hacker News · github.com

OpenAI reduces Codex Model Context Size from 372k to 272k ↗

Community discussions

Hacker News discussion ↗

Historical count: 5 · 5 saved in sample · Historical sample

Analysis ready

Community Sentiment

Negative

Positives

Some users believe that smaller context sizes can lead to better quality responses, suggesting that LLMs get dumber as context stretches too far.
One commenter feels that Codex performs well enough without issues related to context size, indicating improvements in the model's handling of larger projects.

Concerns

Many commenters are frustrated with the reduced context size, arguing that the compaction process results in lost detail and negatively impacts their workflow.
The inability to disable auto-compaction is a major sticking point, with users claiming it leads to hallucinations and unfinished tasks when the context window shrinks unexpectedly.
A frequent sentiment is that the context size reduction feels like a cost-saving measure rather than a genuine improvement, leaving users concerned about its practical implications.