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

	OpenAI reduces Codex Model Context Size from 372k to 272k(github.com)
	367 points by AmazingTurtle 98 days ago \| 163 comments
	tl;dr: Summary not available
	HN Discussion: ↓Reduced context size is a dealbreaker; long context is essential for detailed work and papers ↑Large contexts degrade model quality anyway, so the reduction is fine or even beneficial ~Compaction is lossy and problematic; better to manually manage context via chunking or clearing ↑Codex's compaction works seamlessly in practice, so the reduction isn't noticeable ↓OpenAI should adopt cheaper KV cache techniques (like DeepSeek) rather than shrinking context