Meta releases 816B Muse Spark Open, now the top US open-weights model at AA Index 39
ArtificialAnlys · x · 2026-10-16
Artificial Analysis benchmarked Meta's newly released Muse Spark Open, an 816B-total/42B-active MoE open-weights model that scores 39 on the AA Intelligence Index — the #1 US open-weights model, within a point of Meta's August flagship Muse Spark 1.2 (40).
Key facts:
- Benchmarks: 13 points above NVIDIA Nemotron 3.5 Super VL (26); leads all open-weights models on AA-Omniscience (25) with the lowest hallucination rate (24%) among open-weights models scoring 30+
- Pricing: $1.25/$4.25 per 1M input/output tokens ($0.15 cached) via Together AI, Fireworks, Baseten — $0.84 per Intelligence Index task, under half of GLM-5.3 ($2.01) but 3x DeepSeek V4.1 Flash ($0.27)
- Architecture: 1M context, 62 layers with GQA (16 KV heads), 8 of 256 experts active per token, 1.8B vision encoder; accepts text, image, video, audio
- Deployment: BF16 weights 1.6TB; NVFP4 checkpoint 530GB, runs on 4 Blackwell Ultra GPUs, 8 H200s, or two 512GB M5 Ultra Mac Studios via MLX
- Weakness: ranks last among top open-weights models on the index's four agentic evals — GLM-5.3's entire 6-point lead comes from agentic tasks
- License: custom Muse Model License requiring "Built with Muse Spark Open" attribution; MaaS businesses over $20M annual revenue need a separate Meta agreement
More from Models
- GLM 5.5 rumored to drop soon, coding ability expected to match frontier models — bindureddy · 2026-10-16
- Five rounds of adversarial prompting make Opus 5.5 and Argon converge on identical output — LanceCharger · 2026-10-16
- DIY benchmark ranks 18 models on subjective judgments, GLM-5.3 tops consensus match — arkuto · 2026-10-16
- Bindu Reddy: Google's Argon excels at complex doc writing at half the cost of Opus — bindureddy · 2026-10-16
- MAI-2 is 2x bigger than V4-Pro while V4.1F is 3x smaller: Microsoft's scaling ladder questioned — teortaxesTex · 2026-10-16
- Step 5 Preview hits DeepInfra day-zero: 600B MoE, 1M context, $1/M input tokens — StepFun_ai · 2026-10-16