Rumour: a leak says OpenAI's next big model ships in August. There are two weeks left and OpenAI has confirmed none of it.

This is an unverified rumour, published as one. Since mid-July a leak has circulated claiming OpenAI is preparing GPT-5.7 as an interim flagship ahead of a delayed GPT-6: a new pre-training foundation, a context window above 1.5 million tokens, a training run described as '10-trillion-scale', and a launch inside August. It spread through AI accounts on X and was picked up by several outlets. None of it is confirmed. What OpenAI has actually said is narrower: GPT-5.6 was announced on 9 July in Sol, Terra and Luna tiers, and on 1 August the company named its next major model Astra — without a date, a price, or any statement on whether Astra carries the GPT-6 label. There is no official material for a GPT-5.7 at all. The claim traces back to a single mid-July aggregator article rather than a document, a named source or a leaked build. (Update, 19 August: on 18 August OpenAI published its own account of how it is pacing development. Its largest planned frontier training run is still on hold, and some Astra training and evaluation workloads remain paused until they meet new security requirements. That is not a formal denial of the rumour, but a company that has halted its biggest training run is not two weeks away from shipping a new flagship. Treat the August launch window as effectively closed.)

⚠ Unverified rumor — treat with caution

Related stories