Meta open-sources Muse Spark, topping Kimi and GLM on most agentic and multimodal benchmarks
bindureddy · x · 2026-10-16
Meta has open-sourced its multimodal model Muse Spark under the Muse Open License. Sharath Bulusu-tier commentator bindureddy calls it the best US open-source model, beating Kimi and GLM, and a strong base for custom fine-tunes as Meta races to catch up.
The accompanying benchmark table shows Muse Spark (xhigh) leading on JobBench (61.8), MCP-Atlas (89.2), HLE-Diamond (26.1), AA-Omniscience (25) and ZeroBench (50.0), though Kimi K3 still edges it out on coding benchmarks like Terminal-Bench 2.1 (85.0 vs 76.4) and GLM 5.3 scores higher on GDPval-AA v2.
More from Models
- Muse Code 1.4.8 cuts token usage; all user quotas reset immediately — alexandr_wang · 2026-10-16
- Cornell and Amazon revisit the 16x pretraining compute gap: diffusion LMs need only 3-8x — LucaAmb · 2026-10-16
- Image Decision Benchmark JevImageBench hits v0.3.0 with reshuffled Top 5 — airesearch12 · 2026-10-16
- Forecast: Meta's Muse Spark Open May Become the Best Non-Chinese Open-Weight LLM on Agentic Benchmarks — zsakib_ · 2026-10-16
- Claude Opus 5.5 effort levels tested: low reasoning iterates 9x faster than max — tom_doerr · 2026-10-16
- Matt Turck podcast unpacks Jev: TypeSafe's machine-native AI and its RLHF critique — mattturck · 2026-10-16