Anthropic dropped Claude Fable 5 on June 9th. No preview period, no waitlist.
The headline benchmark: Stripe handed it a 50-million-line Ruby codebase for a migration that would take their whole team two months. Fable 5 finished it in one day.
What Mythos-Class Means
Fable 5 sits above Opus in Anthropic lineup. Mythos 5 is the same base model but for vetted life science and cybersecurity partners.
Fable 5 ships with three safety classifiers (offensive cybersecurity, bio/chem, model distillation). Trigger rate: under 5% of sessions.
Key Benchmarks
- FrontierCode (Cognition): highest score among frontier models
- Hebbia financial reasoning: first place
- Replit vibe coding: wins while using fewer tokens
- Vision: beat Pokemon FireRed using only visual input
Pricing
- Input: $10/M tokens
- Output: $50/M tokens
- API:
claude-fable-5 - Pro/Max/Team/Enterprise: free through June 22nd
Fable 5 targets tasks requiring sustained attention and multi-step reasoning across long contexts. Code migration. Financial analysis. Scientific research. Not demos anymore.
Full analysis: wdsega.github.io
Top comments (0)