Commit

922f1b4

Anthropic just launched Claude Fable 5.2 and Claude Mythos 5.2. Under the hood it's the same model. The difference is that Mythos has fewer safety guardrails and is only available to verified cybersecurity and life sciences organizations. Fable is already available on all paid plans, in Claude Code and through the API. The numbers that caught my attention: - It's 20% cheaper than the previous version, at $8 per million input tokens and $40 per million output tokens. - It generates responses up to 30% faster. - In coding, on "high" effort, it matches Opus 5.5 on max while spending roughly half per task. - 1,000 of its agents formalized the proof of Fermat's Last Theorem in 30 hours, while Opus 5.5 got through 26%. But there's an interesting detail: Anthropic itself recommends sticking with Opus 5.5 as the daily driver and saving Fable for heavier, open-ended technical work. So I'm still not sure if it's worth switching everything over or if it's more for when Opus gets stuck. Every launch comes with pretty benchmarks. What matters is how it holds up in real use, in day-to-day code. Has anyone tried it yet? Full announcement: [www.anthropic.com](https://www.anthropic.com/claude-fable-and-mythos-5-2)
anthropic.comHomeAnthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.

commit 922f1b4