Simon Willison Releases llm-anthropic 0.31 with Fable 5.2 and Mythos 5.2 Support
Simon Willison has released version 0.31 of llm-anthropic, adding support for Anthropic's newly launched Fable 5.2 and Mythos 5.2 models. Developers can now run prompts and test different reasoning levels for these new models directly through the command-line tool. This update allows developers to immediately integrate and experiment with Anthropic's latest high-performance reasoning models using a familiar open-source CLI utility. It highlights how fast developer tooling adapts to major foundational model releases. Testing showed that while lower reasoning tiers of Fable 5.2 work well, the maximum reasoning effort level ('max') failed due to a token exhaustion bug after 128,000 thinking tokens. This mirrors a similar over-thinking bug previously observed in the Opus 5.5 release.
## BACKGROUND
The `llm` ecosystem created by Simon Willison is a popular Python-based utility and plugin architecture for interacting with various Large Language Models from the command line. Anthropic's Fable 5.2 and Mythos 5.2 are advanced configurations of a new large language model designed for demanding technical challenges and long-horizon agentic work.