Meta's 816B Muse Spark Open scores 39 on Artificial Analysis, the 180B Qwen3.8-Flash-Next scores 40. Score vs size for 11 open-weight models
Made this from the Artificial Analysis model pages after Muse Spark Open and Beam got their scores last week.
Proof1 community thread
Key points
- Muse Spark Open (816B, 42B active) scores 39.
- Qwen3.8-Flash-Next (180B, 6B active) scores 40.
- The Unsloth UD-Q4KXL GGUFs are about 506 GB for Muse Spark Open and 111 GB for Qwen3.8-Flash-Next.
- Meta's recommended local setup for Muse Spark Open is two M5 Ultra Mac Studios with 512 GB each.
Sources (1)
- [1]Meta's 816B Muse Spark Open scores 39 on Artificial Analysis, the 180B Qwen3.8-Flash-Next scores 40. Score vs size for 11 open-weight modelsr/LocalLLaMA (top, daily) · Oct 22, 08:14 PM
Made this from the Artificial Analysis model pages after Muse Spark Open and Beam got their scores last week.
- Muse Spark Open (816B, 42B active) scores 39.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 22, 2026Q4 flash next for £300/600 👀 - snapdragon mini PC results update - 315/43 Tok/s A3b, 46/20 Tok/s Flash Next (read/gen) 💥
- Oct 22, 2026capy-flash v0.7.1: Qwen3.8-Flash-Next (125B MoE) now runs in 64 GB of RAM (234 tok/s on two 16 GB cards), and Qwen3.8-27B 3-bit fits on a single 16 GB card (174 tok/s)
- Oct 21, 2026Claude’s personal agent
- Oct 20, 2026Latest open artifacts (#25): DeepSeek V4.1 Flash 1017, American-made models, and new regulatory controls in the US & China
- Oct 20, 2026Hardly any benchmarks for Intel B65, so I got 4x and tested them. Long Context Sweeps for Flash-Next and other models
- Oct 20, 2026How I built my ‘everything’ app
