Codex Users Hit by Tighter Rate Limits and Suspended Pro Subscriptions
From September 13 to 15, multiple developers reported OpenAI Codex capacity and subscription problems, pointing to strained inference compute and limited subscription supply. OpenAI has not officially responded; the compute shortage remains user speculation, but paying users unable to renew have already seen their workflows disrupted.
Confirmed
- On September 13, developer andimarafioti said his 6-month Codex Pro grant ran out just as OpenAI paused new Pro subscriptions, leaving him unable to pay his way back in; he relies heavily on Codex and has consumed roughly 10 billion tokens. MaziyarPanahi relayed the same report, and another user reported similar restrictions.
- On September 14, user markk posted that a single message can ruin a Codex user's day, with a screenshot corroborating the capacity issue, echoing recent complaints about rate limiting and queuing.
- On September 15, developer cocktailpeanut complained that Codex rate limits keep tightening, with the Sol-medium model draining account quota quickly, arguing the problem is not a specific new model (Astra) but a possible sudden compute shortage across OpenAI.
- The same day, developer lxfater said upgrading a Codex subscription was not possible, joking that good things sell out, with commenters discussing whether they faced the same limit.
Unconfirmed
- OpenAI has not explained the compute shortage or how long the Pro subscription pause will last; strained capacity remains speculation from users such as cocktailpeanut.
Why it matters
- Codex has become a core productivity tool for developers like andimarafioti, so quota cutoffs directly disrupt work; the inability to pay for renewal exposes a tension between subscription supply and compute capacity, and if the shortage is real, it may pressure OpenAI's product experience and commercialization going forward.
2026-09-13 ~ 2026-09-15 · 5 related posts
Primary sources
- [source] OpenAI pauses Pro subscriptions, leaving a Codex power user stranded after 10B tokens — andimarafioti · 2026-09-13
- [source] Codex Users Report Capacity Woes as OpenAI's Compute Strain Looks Real — mark_k · 2026-09-14
- [source] Dev: Codex rate limits tightened hard, hints OpenAI is running out of compute — cocktailpeanut · 2026-09-15
- Codex demand so hot users say they can't even upgrade their plans — lxfater · 2026-09-15
1 near-duplicate retellings: MaziyarPanahi