OpenAI reduces Codex Model Context Size from 372k to 272k tokens; community builds Codex Resets tracker

What happened

On July 18, 2026, OpenAI released Codex 0.144.6, which corrected the GPT-5.6 Sol, Terra, and Luna context windows to 272,000 tokens — tacitly confirming the previously advertised 372,000 was an error. Simultaneously, the developer community launched Codex Resets (codex-resets.com) tracking unpredictable usage-limit restorations.

Context and impact

For developers running production agents in Codex, this is a double problem: not only a context reduction, but also chronic unpredictability of usage limits. The tracker documents 35 resets over 26 weeks with an average interval of 8.9 days — with no published schedule.

Details

• Codex 0.144.6 (July 18, 2026): context window corrected to 272k input + 128k output tokens
• The previously advertised 372k tokens was never publicly explained
• Codex Resets tracker: 35 documented resets, 8.9-day average interval, 67.7-day max gap
• Resets are announced via tweet, not via changelog
• HN discussion on context window: 46 points; Codex Resets discussion: 202 points