New Plugin Lets You Use Meta’s Muse Wonder AI Models
Simon Willison updated his llm-meta-ai tool to work with Meta’s latest top-tier reasoning models.

A program add-on that lets you test Meta's new AI reasoning and image models through simple commands without writing custom code.
- Install the plugin on your command line tool
- Pick a reasoning level from six options
- Ask the model to generate something
- Compare results across different reasoning levels
Plugin updated for Meta’s latest models
Simon Willison released version 0.2 of his llm-meta-ai plugin, which lets developers and researchers access Meta’s newest artificial intelligence models through a familiar command-line interface. The update adds support for Muse Wonder, a pair of new models Meta recently announced as its most capable yet.
Muse Wonder is available in a standard version and a heavier Muse Wonder Full version. Early benchmarks from Artificial Analysis suggest the Max setting on Muse Wonder may match the performance of GPT-6.1 Sol, placing it in what Meta now calls its "Fable class" of frontier models.
Testing the models with pelicans
To test how the models handle reasoning at different levels, Willison asked each one to generate images of pelicans across six reasoning tiers — minimal, low, medium, high, xhigh, and max. He shared the results publicly so others can compare how the output quality changes as reasoning effort increases.
He also compared the new Muse Wonder results against earlier versions: Muse 1.3, released in early September, and Muse 1.2, which came out in early August. Willison noted that Meta has been shipping new model versions roughly once a month, and the improvements between releases are noticeable even in simple image-generation tests.
Meta ships new AI models roughly once a month
- Early August 2026Muse 1.2 released
- Early September 2026Muse 1.3 released
- October 2026Muse Wonder announced as Meta's most capable models yet
Dates and model names from Simon Willison's blog post, October 12 2026
What the plugin does
The llm-meta-ai plugin acts as a bridge between Willison’s LLM command-line tool and Meta’s AI API. It lets users call Meta’s models without writing custom code, making it easier to experiment with different reasoning levels and compare outputs side by side.
Willison’s blog post, published on October 12, 2026, framed the update as a practical way for the open AI community to keep pace with Meta’s rapid release schedule. Rather than waiting for formal reviews, developers can now try the models themselves and form their own conclusions about what the latest generation of reasoning AI can do.
Corrections and updates
- Corrected on 21 October 2026. An earlier version said Meta is competing with OpenAI and Google, but none of the provided sources mention either OpenAI or Google at all. The sources only discuss Meta's own AI divisions, products (Muse, Llama), and the acquisition of Scale AI. An earlier version said Willison asked models to draw pelicans riding bicycles; the source only mentions drawing pelicans with no bicycles. It also stated models were released monthly since August 2026, but the source references August 2025 as the starting point.
Key facts
- llm-meta-ai 0.2 adds support for Meta's Muse Wonder and Muse Wonder Full models
- Artificial Analysis scores Muse Wonder Max as roughly equivalent to GPT-6.1 Sol
- - Willison tested six reasoning levels by asking models to draw pelicans
- - Meta has released new Muse models approximately monthly since August 2025
Sources
Every fact in this story comes from the source above.
If this story concerns you or your company, or uses something that belongs to you, ask for a correction or for it to be taken down. How requests work
Keep reading
New modelsMeta's Muse Wonder Brings It Back to the AI FrontierHerald of AI
Meta's Muse Wonder Brings It Back to the AI Frontier
Meta's newest models rival the best from OpenAI and Anthropic, marking a major comeback after years of falling behind.
New modelsMeta's Muse Ember matches Claude Haiku at a third of the costHerald of AI
Meta's Muse Ember matches Claude Haiku at a third of the cost
Meta's smallest AI model does routine agent tasks for about a third of what rivals charge.
New modelsSimon Willison Updates llm-gemini to Support New Gemini 4 Argon ModelHerald of AI
Simon Willison Updates llm-gemini to Support New Gemini 4 Argon Model
The developer behind the popular llm plugin adds Google's latest low-cost model, which promises strong performance at a fraction of rival prices — at least until December.