Claude Fable 5 and Mythos 5

Anthropic shipped Claude Fable 5: the first Mythos-class model made generally available, two months after the Mythos leak we covered in April. Alongside it, Claude Mythos 5: the same underlying model with some safeguards lifted, deployed only to ~50 vetted cyberdefenders and infrastructure providers through Project Glasswing.

• Pricing: $10 / $50 per million tokens, included on paid plans until June 22, then it eats usage credits
• Benchmarks: 80.3% on SWE-bench Pro (Opus 4.8: 69.2%, GPT-5.5: 58.6%), 29.3% on FrontierCode Diamond and a new Humanity's Last Exam high (53.3%). Debuted #1 in LMArena's Code Arena with a 98-point lead.
• Safety classifiers: high-risk cyber and bio queries reroute to Opus 4.8, triggering in under 5% of sessions
• System prompt: ~120,000 characters, leaked by Pliny within two days. Of course.

The model is genuinely good, the first draft of this blog post was made by Fable 5 itself (spawning something like 200 sub-agents and token-maxxing two consecutive usage windows).

Sources: official announcement, TechCrunch, CNBC