GPT vs Claude: Context Window Strategy Divergence

PawelHuryn · x · 2026-07-20

Hacker News discovered that GPT-5.6 Sol's max context window in Codex was quietly reduced by 100k tokens to 272k tokens, labeled as a bug fix in release notes.

The author finds this puzzling because Anthropic took the opposite approach: retaining a 1 million context window in Cowork without a /compact command, instead restricting the model itself.

Related event: OpenAI Quietly Reduces Codex Context Window to 272k Tokens(5 posts)→

Original post →

More from Models

Models channel →