🎨 Claude Fable 5.1: one pelican — from 10 cents to $3.30
Simon Willison tested Anthropic's new model on the 'pelican on a bicycle' test at all reasoning levels — from low to max, reasoning cannot be disabled. The best pelican is on max; at the request of a Hacker News reader, Willison animated it with the same model for $1.37.
🌍 Reasoning effort has become the main price and quality knob: low — 23.8s and ~10 cents, xhigh — 7m 51s and $1.83, max — 13m 54s and $3.30. On the Terminal-Bench-Science 0.1 benchmark, the model scores 52.6% compared to 29.0% for Opus 5.
👤 All pelicans with reasoning transcripts are in the author's gist, and the animation can be reproduced via llm-CLI (llm -m claude-fable-5.1). Budget: a complex request on max will cost more than $3 per run, on low and medium — cents. There are no reasoning traces on low and medium — a mystery, writes the author.
Source 1: https://simonwillison.net/2026/Sep/1/claude-fable-5-1/ Source 2: https://news.ycombinator.com/item?id=49530472