Fable 5.2 and Mythos 5.2 arrive in llm-anthropic 0.31
Updated
Illustrative photo
Two new models, one install command
Anthropic released Fable 5.2 and Mythos 5.2 on November 9, 2026. Simon Willison added both to llm-anthropic, the plugin that gives the llm command-line tool access to Anthropic's Claude models, and shipped it as version 0.31 the same day.
Getting started takes three commands: install llm, install llm-anthropic, then set an Anthropic key. After that you point the tool at the model with -m claude-fable-5.2 and pass a prompt. Willison's example asks for an SVG of a pelican riding a bicycle.
The max reasoning level ran out of tokens
Willison ran the pelican prompt at all five reasoning levels: low, medium, high, xhigh and max. The best image, in his view, came from xhigh. At max, the model returned no image at all and exhausted its tokens after spending 128,000 thinking.
He writes that he saw the same bug with Opus 5.5, but that it is new for the Fable series: both Fable 5 and Fable 5.1 produced max-effort pelicans without trouble.
What to check before you switch
If you rely on the top reasoning setting for image output, the xhigh level is the one that worked in this run. The max setting is where the failure showed up, and it is the setting Willison flags as a regression for Fable. I haven't run these models myself, so the token count and the image results are his, not mine.
The practical step is to try your own prompt at max before you build a workflow around it.
Affiliate links: we may earn a commission from qualifying purchases, and it never changes our data or rankings. They open Amazon's search results: prices and availability are shown there.