OpenAI Cancels GPT-6.1 Astra Release Over Safety Alignment Regression

According to an exclusive Wall Street Journal report, OpenAI has canceled—rather than delayed—the planned October launch of its next-generation model GPT-6.1 Astra in ChatGPT and Codex, after internal safety testing found regressions in two alignment metrics. It is a rare case of a frontier model being scrapped for safety reasons, and Polymarket has already opened prediction markets on its eventual release timing.

Confirmed

Unconfirmed

Why it matters

2026-09-29 ~ 2026-09-30 · 42 related posts

Full story(8 episodes)→

Primary sources

20 near-duplicate retellings: mark_k · ns123abc · CucumberAccording813 · nordicinst · BorisMPower · ChrisGPT · thesaraharminta · haydenfield · 智东西 · kimmonismus · kimmonismus · YashasGunderia · 新智元 · ShakeelHashim · Next_Tower5452 · emmanuelvivier · emmanuelvivier · dotey · wellecks · TansuYegen