Why Claire Vo Ran Meta's Muse Code Headless Inside Claude Code
Claire Vo explains the How I AI benchmark methodology and why she used Claude Code to run Meta's Muse Code headlessly.
Is Meta’s Muse Wonder model any good?
With Claire Vo · Sunday, October 18, 2026
Claire Vo runs Meta's new frontier AI models, Muse Wonder and Wonderful, through the How I AI benchmark across coding, agentic workflows, prototyping, and creative tasks. She finds the model exceptionally capable at agentic computer use, email writing, and voice orchestration, while noting its active focus mode aids long-running tasks. However, Muse falls significantly behind competitors like Claude Opus 5.5 and Astra in front-end design, UI aesthetics, SVG generation, and video editing.
Key takeaways
- The How I AI benchmark evaluates frontier models on roughly 20 practical workflows divided into five buckets: prototypes, agentic tasks, coding, writing, and creative output. Read more →
- Meta's Muse Wonder swept all four agentic evaluations in the How I AI benchmark, outperforming rival models in back-end coding, inbox triage, and computer use. Read more →
- Claire Vo benchmarked Meta Muse Wonder against Claude Opus 5.5 and Astra, finding its front-end design quality trailing competitors by 12 to 18 months. Read more →
- Mark Zuckerberg framed Meta's Muse Wonder and Wonderful as direct competitors to GPT-6, Astra, Fable, and Claude Opus 5.5, with heavy emphasis on law, finance, and competitive coding on Deep Sue. Read more →
- Claire Vo tested Meta Muse Wonder and Wonderful on the How I AI benchmark, uncovering an operational split between instant single-prompt tasks and prolonged multi-step workflows. Read more →
5 articles from this episode
Meta Muse Wonder Beats Claude on Agentic SaaS Workflows
Claire Vo benchmarks Meta Muse Wonder against Claude across complex SaaS admin workflows, computer use, and adversarial prompt injections.
Why Meta Muse Wonder Fails at Front-End UI Design
Claire Vo tested Meta Muse Wonder against Claude Opus 5.5 and Astra. Here is why the model fails at front-end UI and web design.
Benchmarking Meta Muse Wonder: Strong Agents, Weak Front-Ends
Claire Vo benchmarks Meta's Muse Wonder against Claude Opus 5.5 and Astra, revealing strong agentic execution paired with lagging UI design.
Meta Muse Splits Speed: Fast on Voice, Slow on Agentic Work
Claire Vo benchmarks Meta Muse Wonder: instant on voice and one-shot tasks, but slow on multi-step reasoning like inbox triage and back-end code.
More How I AI episodes
- Meta Muse: the AI agent people actually loveOct 18 · 5 articles
- How the OpenAI team uses ChatGPT Sites dailyOct 11 · 5 articles
- I’m using Jev more than Opus 5.5 or GPT-6. Here’s why.Oct 4 · 5 articles
- Jev: 8 real use cases this fast, cheap modelOct 4 · 6 articles