https://www.anthropic.com/claude-fable-and-mythos-5-2
What changed
- writes more clearly, token generation is 30% faster
- 20% cheaper than Claude 5.1 — $8 per million input tokens, $40 per million output tokens
- at double the price, a fast mode is available with generation 2.5x faster
- Anthropic gives several examples of orchestrating 1000 agents to solve tasks — for example, writing an operating system or [formalizing a proof of Fermat's Last Theorem](https://claude.dev/blog/safely-scaling-multi-agent-teams-with-fable-5-2/)
- accounts with Fable access will get 1 to 3 resets today, and the quota limit is also up 50%
- the most notable benchmark result is over 40% on the Remote Labor Index (RLI), a test that measures what share of real remote work a model can complete on its own
Why it matters
Fable and Mythos 5.2 are launching amid a pricing race: OpenAI recently released GPT-6, Google shipped Gemini 4 Carbon. Cutting prices while speeding up models is how Anthropic keeps developers on board while it prepares for an IPO, where revenue growth will directly shape the company's valuation.