This week The Information reported that OpenAI could drop a new Astra, a new Sol and a cybersecurity model as soon as Friday, which is today. The leak-watchers call them GPT-6.2 Astra and Sol, though OpenAI hasn’t announced either. There was never a 6.1 Astra, because OpenAI scrapped it over safety concerns. ArtificialWatch’s release log puts OpenAI’s median gap between frontier models at 41 days, and 41 days after GPT-6.1 Sol was 9 November. My call is GPT-6.2 Astra and Sol before 13 November.

Anthropic’s median frontier gap is 22 days, and it’s already 39 days since Claude Sonnet 5.5, so the median is no use here. That figure comes from a burst across tiers. The tier that’s due is the top one, and it runs slower: Fable 5.1 came 84 days after Fable 5, and 84 days from 1 September is 24 November. Opus 5.5 also performs at the level of Fable 5.1 on most work, by Anthropic’s own account, and a flagship the tier below has caught isn’t much of a flagship. So the next Fable model lands in the last week of November. LLM Countdowns says 30 November, close enough that I won’t argue.

The trackers have been wrong already this autumn. Forkast counted projections for Anthropic and OpenAI on 28 and 29 October, and neither lab shipped a frontier model that week. I got a Thursday wrong myself in September.

Sources: