I’m a bit giddy about Fable, and I may be recanting my statements from this morning. It just managed to break through a bottleneck that neither Opus 4.8 nor Codex has accomplished in a week.
Ten days ago I was saying Claude had been relegated to double-checking ChatGPT, and that I wasn’t sure 4.8 was any better than 4.7. Today it feels like I’m back in the glorious early 4.6 days, when anything was possible. Only this time the feeling comes with a leap ahead attached.
Some context for anyone who hasn’t refreshed yet. Fable 5 is the model Anthropic released this morning, and the company describes it as “a Mythos-class model that we’ve made safe for general use”. Anthropic says it “can work autonomously for longer than any previous Claude models,” and that the longer and more complex the task, the larger its lead over the rest of the Claude line. From today through June 22 it’s included on Pro, Max, Team and seat-based Enterprise plans at no extra cost. On the API it costs $10 per million input tokens and $50 per million output, which TechCrunch notes is double the price of Opus 4.8.
Still, a bottleneck that held for a week against two capable models and then fell is not the kind of evidence a benchmark gives you.
I guess I got lucky with early usage of Fable too.
Now the meter, because there’s a wrinkle. I’d otherwise have been grateful for the full reset, but my usage per minute was roughly twice as fast after the reset, using Fable. My guess is they forgot to tether the new model to 2X usage costs. Whatever the mechanism, the API price is double Opus 4.8’s, and the subscription meter seems to agree with the API. Fable, it turns out, is voracious at the meter. If you plan to lean on it before June 22, watch the usage screen as closely as the calendar.
So the “giveaway” is really more of a takeaway reset.
Has anyone else refreshed their Claude setup today to see whether it’s in their subscription yet?
