On August 7, 2026, Mark Zuckerberg announced that Muse Code — Meta's terminal coding agent, built in the mold of Anthropic's Claude Code — was entering public beta, powered by Muse Spark 1.2, Meta's third model release in four months. The host, developer behind T3 Code, conducted a live, 44-minute stress test including codebase audits, game remakes, and a 222-pull-request audit. Muse Spark 1.2 scores 54 on the Artificial Analysis Intelligence Index, up 11 points from April's 1.0 release, effectively tied with GPT-5.5 and Grok 4.5, and narrowly behind Opus 5, Fable 5, 56 Soul, and Kimi K3. Meta's benchmark charts omitted its two strongest competitors. The contributor tier is priced 10 to 20 times below the standard API. At that price, the model audited 222 pull requests in under five minutes for 10 cents, with a cleaner report than usual tools. At its deepest test, it hallucinated an entire wrong product, produced a plan that OpenAI's 56 Soul scored 4.8/10 against Fable's 7/10, and the code did not work. The host concludes it is a fast, cheap model with aesthetic personality that "knows how to act like a modern smart model, but doesn't know what the modern smart models know." Pricing