Version 0.2 of the llm-meta-ai plugin adds Muse Wonder access and shows one image prompt across six available reasoning levels.
Simon Willison released llm-meta-ai 0.2 on October 12, updating his plugin for the Meta AI API to work with the new Muse Wonder model. The release gives users of Willison’s LLM tooling a way to access that model through the existing plugin rather than requiring a separate integration.
The Meta AI release referenced by Willison introduced Muse Wonder and Muse Wonder Full, describing them as the company’s new “most capable models.” His post does not provide technical specifications, API prices, usage limits or availability details. It also says only that the plugin was updated for “the new model,” so the announcement does not establish support for Muse Wonder Full in version 0.2.
To demonstrate Muse Wonder, Willison used his recurring prompt involving a pelican riding a bicycle. He generated examples at six reasoning levels: minimal, low, medium, high, xhigh and max, and displayed the max-level result in the release post. These images document how one prompt was handled under the six settings; the post does not present them as a controlled assessment of quality, latency or cost.
Willison also linked the demonstration to earlier pelican images from Muse 1.3, released on September 2, and Muse 1.2, released on August 5. He judged the improvements between successive models to be material, at least for drawing pelicans. That qualification is important: the post offers a narrow visual comparison and does not claim that the same degree of improvement applies to other prompts or tasks.
The immediate practical consequence is that Muse Wonder can be used through an updated llm-meta-ai plugin, with six reasoning settings illustrated by the author. Choosing among those levels for production work would still require separate evaluation, because this demonstration supplies no measurements for response time, spending, reliability or consistency. It is therefore useful as evidence of plugin support and available settings, but not as a benchmark for selecting one mode over another.
Sources
Originally published by AI Feed: Simon Willison Adds Muse Wonder to llm-meta-ai 0.2.
Top comments (0)