Codex issue: 5-hour window drained to 0% by just two requests in 20 minutes
BlueJ-D · ghdev · 2026-09-16
GitHub issue #45828 reports Codex CLI (0.154.0, Plus plan, GPT-6, Linux) hitting a repeat of the Sept 9 anomaly: /status showed 100% allowance at session start (1:24am, renewing 6:24am), yet two task requests within 20 minutes drained the 5-hour window to 0%.
The user notes they had used gpt-6-astra high all day to build a plan file for a multi-task goal, then switched to the medium model to execute stage 1 — questioning how stepping down to medium could consume more, suspecting either a glitch or OpenAI quietly shaving the allowance daily. The actual work was tiny: the last commit was under 100 lines, and the interrupted run touched 15 files with one-to-two-line changes plus one new 30-line file. The /usage chart also shows today's square notably dimmer than the past month. A possible repro: build the plan with astra high, switch to medium, and start the task list; yesterday's similar but larger workload never came close to the limit.
More from Models
- LAB legal benchmark flaw: case docs leak planted issues directly to models — andersonbcdefg · 2026-09-17
- MiMo eval chart shows judge and probe disagree 60% of the time, sparking reward-hacking concerns — andrew_n_carr · 2026-09-17
- Fans mourn the end of Codex lead's regular daily reset cadence — kimmonismus · 2026-09-17
- Users report Google Astra feels noticeably degraded over past two days — pwlot · 2026-09-17
- OpenAI burns 20% as much compute on monitoring as the model itself, SemiAnalysis says — kevinnbass · 2026-09-17
- Google releases Gemma 3n: 2GB RAM multimodal model, first sub-10B to top 1300 on LMArena — joemeno · 2026-09-17