brandonmusic/Muse-Spark-Open-816B-A42B-TrellisMX-K3
archived · low-signalmodel
A 3-bit TrellisMX build of meta-models/Muse-Spark-Open-816B-A42B (BF16 base, revision 0fb02478b02565b2ce7fc57e6051c91172774ac9) sized for four 96 GB RTX PRO 6000 Blackwell cards at tensor parallel 4, with the image, video and audio towers kept. 24B params. License: muse-model-license. Vendor-reported: "Intervals are window-block bootstrap (5,000 resamples, BCa) and Wilson for top-1." Vendor-reported: "All 128 windows, without the overlap screen: 0.0333 [0.0283, 0.0399], top-1 94.52% (227,038 text rows)."
- Vendor-reported: Intervals are window-block bootstrap (5,000 resamples, BCa) and Wilson for top-1.
- Vendor-reported: All 128 windows, without the overlap screen: 0.0333 [0.0283, 0.0399], top-1 94.52% (227,038 text rows).
- Org
- brandonmusic
- Released
- 2026-10-16
- Status
- archived · low-signal
- License
- muse-model-license (restricted)
- Parameters
- 23.6B
- Task
- image-text-to-text · vision
- Signal
- 4 HF likes · 8 downloads
- Weights
- gated
- Archived
- 2026-10-19 (no longer listed by the collector; stats as last seen)
Runs on
estimate from params; need = weights + FP16 KV cache for 8K tokens + 1GB. A sizing calculation, not tested.
- Q4 ~17.2 GB (estimated size): 24GB GPU (3090 / 4090), 24GB GPU (7900 XTX), Mac 24GB, Mac 36GB, CPU only, 32GB RAM (slow)
- Q8 ~28.1 GB (estimated size): 32GB GPU (5090), 48GB GPU (RTX 6000 Ada / A6000), Arc B70 32GB, Mac 48GB, Mac 64GB (M1 Max)
- FP16 ~50.2 GB (estimated size): 96GB GPU (RTX Pro 6000 Blackwell), 1× DGX Spark (128GB), 2× DGX Spark (256GB), Strix Halo 128GB unified (~96GB usable), Mac 96GB, Mac 128GB, Mac 192GB, Mac 256GB, Mac 512GB, CPU only, 64GB RAM (slow), CPU only, 128GB RAM (slow), CPU only, 256GB RAM (slow)
Links
First published here 2026-10-16 · updated 2026-10-19 · markdown · JSON