OpenAI has fixed three defects behind a week of user complaints that its new GPT-6 Astra model grew dumber after launch, and is resetting usage limits for Codex subscribers.
Key Points:
- Astra users spent the week posting side-by-side tests claiming the model had been quietly weakened since launch.
- OpenAI's Codex lead named three causes, including a context experiment that touched roughly 4,000 to 5,000 users.
- The company's previous flagship, GPT-5.6 Sol, drew the same accusation in July.
OpenAI Fixes Astra Quality Defects
Codex product lead Tibo Sottiaux said the team worked with users to trace the faults, then shipped fixes and a full usage reset before midnight local time, according to an update published Saturday. Skills written for earlier models were triggering too often, sometimes stopping Astra from checking its own work.
An optional context management experiment could also make the model quit early or reply to stale messages, and OpenAI disabled it after roughly 4,000 to 5,000 users were affected by the test. Some engines had been configured incorrectly and measurably degraded quality on the tail traffic routed through them, so the company removed those engines and made several smaller repairs. Sottiaux wrote that the model now tracks a user's latest message more accurately.
Also Read: Ethereum Faces $2,800 Supply Wall After ETF Inflows Jump To $216M
Astra Users Report Coding Regressions
Complaints built through the week on X, where developers ran identical prompts against launch-day Astra and the current version, kept every setting the same, then compared the results. Developer Pranjal Paliwal, who had praised the model days earlier, went back and read the code it wrote for him, then called the outcome a regression rather than progress.
Dax Raad, who builds the coding tool Opencode, moved his team back to the older GPT-5.6 Sol after spending on Astra doubled for downsides that he judged were not worth the money.
One forum poster described Astra as now level with Sol on identical tasks, with the same retries. Not everyone accepted that reading, and a pseudonymous user writing as Antikythera argued the model had always been lazy and fond of bullet points, so users had simply stopped being dazzled.
Compute Limits Shadow GPT-6 Astra
Capacity has shadowed the rollout since the model reached wide availability in early September.
OpenAI has cut Astra usage limits for heavy ChatGPT users, according to user reports, and paused new $200 Pro subscriptions, a step Sottiaux confirmed on Sept. 10. The model still costs $10 per million input tokens and $50 per million output tokens, which is 2.5 times what Sol charged when it launched.
This is the second OpenAI flagship in two months to draw the same accusation from paying users. Users said in July that Sol's top reasoning mode had gone shallow overnight, and Sottiaux denied weakening it deliberately while confirming that the company had been experimenting with reasoning effort.
Read Next: Nvidia's Jensen Huang Says Security Vendors Benefit From AI Threat Panic
