Fable 5.2 and Mythos 5.2: worth moving off Opus 5.5?

Anthropic put out Fable 5.2 and Mythos 5.2 yesterday, and their advice: most people should still start with Opus 5.5, and only move up to Fable for the really open-ended technical work.

Which honestly makes sense. Fable got cheaper, about 20% below 5.1, so $8 per million input tokens and $40 per million output, and it's supposed to be up to 30% faster. But for the integration and automation stuff I care about, the real question is whether the extra spend buys anything on tasks Opus already handles fine.

There's also a fast mode, up to 2.5x quicker at double the price ($16 / $80). Curious who actually needs it. Maybe long agent runs where a person is sitting there waiting on the result?

Mythos 5.2 is the same model with some cybersecurity and biology safeguards taken off, and you have to apply through Anthropic's verification programs to get it. So for most of us it's Fable or nothing.

On their FrontierCode 2.0 eval, Fable 5.2 scored 49.1% at xhigh effort against 45.5% for Opus 5.5 (Fable 5.1 was 32.4%). A real jump from 5.1, smaller gap over Opus, and it's one eval.

If anyone has moved a real workflow over already, I'd like to hear where it made a difference, and where you went back to Opus.

Links: Anthropic's announcement ·