[AINews] Claude Fable/Mythos 5.1: new SOTA model, 75% cache price cut but 70% more output tokens

With Astra clearly finally warming up for a full launch (with @sama and @openai writing about it again after a month of self imposed pacing), there’s a familiar window to take the narrative with the round robin of model launches, with Grok 4.7 and Gemini Flash 3.8 also on the way. But that’s also perhaps not the best way to frame today’s launch… which got well over 12M views updating the sitting world best model yet again: We’re introducing Claude Fable 5.1 and Claude Mythos 5.1. They're the world’s most advanced models for coding and knowledge work. 6:03 PM · Sep 1, 2026 · 12.8M Views While per-token pricing is the same as Fable/Mythos 5, the cache reads had a 75% price cut… great news for long sessions/long context users, however offset by observed 1.7x output token usage increases per Artificial Analysis, for a total net per-task cost increase of 20% (see recap below). Also don’t World Labs’ Astra launch, by far the most impressive world model launch we’ve ever seen, and on a regular day would have easily gotten title story cards. You can catch up on Fei Fei and Justin Johnson’s vision on our pod and trace from…