Can I run Muse-Spark-Open-816B-A42B-GGUF on an NVIDIA B300?
Yes, just
Its UD-IQ2_XXS file (226 GB) fits in the 259 GB an NVIDIA B300 can use, with room for ≈ 495K tokens of context. That file is under 3 bits per weight, so expect noticeably lower quality.
Full analysis of Muse-Spark-Open-816B-A42B-GGUF
Everything that runs on NVIDIA B300
Get the UD-IQ2_XXS file on Hugging Face
llama-server -hf unsloth/Muse-Spark-Open-816B-A42B-GGUF:UD-IQ2_XXSFiles and context on NVIDIA B300
Each file of Muse-Spark-Open-816B-A42B-GGUF (and of GGUF versions the site has analyzed), and how much context fits next to it, with the context cache in 16-bit or 8-bit.
| File | Size | On NVIDIA B300 | Context (16-bit) | Context (8-bit) |
|---|---|---|---|---|
| UD-Q4_K_XL | 506 GB | Doesn't fit | — | — |
| UD-Q3_K_XL | 401 GB | Doesn't fit | — | — |
| UD-IQ3_XXS | 340 GB | Doesn't fit | — | — |
| UD-Q2_K_XL | 311 GB | Doesn't fit | — | — |
| UD-IQ2_S | 275 GB | Doesn't fit | — | — |
| UD-IQ2_XXS | 226 GB | Tight | 495K | 997K |
| UD-IQ1_M | 209 GB | Tight | 743K | 1M |
| UD-IQ1_S | 176 GB | Fits | 1M | 1M |