Anthropic releases Fable 5.2 and Mythos 5.2 models
Anthropic’s latest models raise coding, science and cybersecurity scores while cutting price and adding stricter safeguards against misuse.
What happened
Anthropic released Claude Fable 5.2 and Claude Mythos 5.2, a paired model launch aimed at ambitious technical work. The two systems share the same underlying model, but Fable includes stronger safeguards for general availability while Mythos is reserved for vetted cybersecurity and life-sciences users.
Anthropic reports a 49.1% result for Fable 5.2 on FrontierCode 2.0 at its highest effort setting, compared with 32.4% for Fable 5.1. It also claims improvements in computer use, scientific research, robotics simulation and chart understanding. Mythos 5.2 reportedly found 10.2% more vulnerabilities than Opus 5.5 in Anthropic’s internal CVE Rediscovery evaluation, while its protein-design and cybersecurity scores also improved.
The release is notable for its cost structure. Fable and Mythos 5.2 are priced at $8 per million input tokens and $40 per million output tokens, with an optional fast mode costing twice as much. Anthropic says Fable 5.2 can match Opus 5.5 on some coding tasks at roughly half the task cost.
Why it matters
The launch pushes frontier models toward sustained, tool-using work rather than isolated question answering. Anthropic says Fable 5.2 helped complete large software migrations and solve two open theoretical-physics problems, but those claims remain company-reported and require independent replication. More immediately, the combination of stronger agentic coding, lower prices and broader platform availability could pressure rival labs on both capability and inference economics.
Anthropic also says it reduced data harvested by illicit distillation by 90% to 99% through new protections. That makes the release important beyond benchmarks: it treats model extraction and controlled access as part of the product’s core security design.