Meta's 816B Muse Spark Open lands with Unsloth quants from 176GB and KLD benchmarks
danielhanchen · x · 2026-10-16
Meta released Muse Spark Open (816B-A42B), billed as the most capable US open-weight model to date, with text/image/video/audio input and 1M-token context — runnable on a Mac Studio via dynamic 2-bit Unsloth GGUFs. Unsloth shipped UD-3 quants down to 176GB (UD-IQ1S) and published KLD benchmarks showing General/Chat/Combined divergence steadily dropping from IQ1S through Q4KXL (176–500GB), saying they recovered as much accuracy as possible for smaller quants. Fully multimodal and supported in Unsloth Desktop.
More from Infra
- Copper to optical: optical backplanes seen as necessary as racks move to optics — BenBajarin · 2026-10-16
- Polymarket prices 28% odds a US state enacts data center moratorium by year-end — Polymarket · 2026-10-16
- Pennsylvania residents reject $10,000-per-household offer for massive Amazon data center — Polymarket · 2026-10-16
- PyTorch now runs natively on AWS Trainium with zero code changes — PyTorch · 2026-10-16
- Chinese labs run on ~1/10 US compute; ByteDance pays 2-4x to hoover up SEA supply — robleclerc · 2026-10-16
- Zymtrace raises $12.2M and hires ML engineers to build agents that fix GPU bottlenecks — cfregly · 2026-10-16