Muse Ember debuts: near Spark 1.3 intelligence at 17x lower price, first Wonder Full-trained model
alexandr_wang · x · 2026-11-10
Alexandr Wang's team released Muse Ember, their smallest model yet, scoring 36 on the Artificial Analysis Intelligence Index — nearly matching Muse Spark 1.3 (max)'s 37 — at roughly 17x lower cost per task ($0.14 vs $2.39). Pricing is $0.10/$0.50 per 1M input/output tokens, versus $1.25/$4.25 for Spark 1.3.
It is the first model trained with Muse Wonder Full and is available now via the API and Muse Code.
More from Models
- Nuance Labs sets hard limits on its voice model: no real-person cloning, no AI girlfriends — iamaliveix · 2026-11-10
- One prompt, one playable Harry Potter × Fortnite mashup from Fable 5.2 — mattshumer_ · 2026-11-10
- Ethan Mollick tests Fable 5.2: launches a business, hacks hardware, calls it a Jagged AGI — alexalbert__ · 2026-11-10
- Nuance Labs found pushing benchmarks higher made its model feel worse to talk to — iamaliveix · 2026-11-10
- Claude Fable 5.2 benchmark dump: sweeps 12 evals, robotics jumps to 64.9% — mark_k · 2026-11-10
- SpaceX says Grok hits 99.9% accuracy spotting debris, beating human inspectors at 94% — XFreeze · 2026-11-10