Scale AI's Muse, built with Meta's Muse Spark 1.3, tops the App Store and PRBench
alexandr_wang · x · 2026-09-23
Scale AI CEO Alexandr Wang announced Muse, an agent built in partnership with Meta that has hit #1 on the App Store. The underlying model is Meta's Muse Spark 1.3, with Scale's team collaborating closely on evaluations, training data, and testing.
- Muse excels at tool use, multi-turn conversation, tutoring, professional reasoning, and end-to-end software engineering—tasks that demand holding context over long-horizon workflows, judgment under uncertainty, and recovery from partial failures.
- On PRBench, where domain experts write tasks scored on rubrics rewarding professional utility, Muse has held the top spot in finance and law since version 1.1.
More from Models
- One RL Run Cost 130 Hours, 75B Tokens and $2.6M — the Real Price of Scaling RL — burny_tech · 2026-09-23
- Qwen 3.8 27B local coding session runs 3 days on one RTX 4090, then spews endless slashes — Tiny-Entertainer-346 · 2026-09-23
- Claude adds banked usage limit reset button on web and desktop — airesearch12 · 2026-09-23
- Third-party test: Claude Opus 5.5 renders finer 3D scenes but costs 13x more than GPT-6 Sol — testingcatalog · 2026-09-23
- Tester claims Claude Opus 5.5 has the best visual design output of any model tested — burny_tech · 2026-09-23
- GPT-6 Sol Codex system prompt leaked: over 294,000 characters dumped on GitHub — gaganghotra_ · 2026-09-23