Hacker News
OpenAI Quietly Cut Codex's GPT-5.6 Context From 372K to 272K — Exactly at the 2x Pricing Boundary
A merged Codex PR (openai/codex#33972) shipped July 19 caps the effective GPT-5.6 window at 272,000 input tokens plus 128K reserved output, down from 372K, and landed only as model metadata — no blog post, deprecation notice, or migration guide. The number is not arbitrary: OpenAI prices prompts above 272K input tokens at 2x input and 1.5x output for the full request, so Codex users can no longer opt into the published long-context tier at all. 356 points and 161 comments on HN, with one engineer reporting $400 burned in an afternoon because their harness kept feeding repo context past the boundary and Cursor/Windsurf don't surface the new limit.
↳ Follow the thread