Muse Spark Open (816B-A42B)
Files and memory at 8K context
| Quant | File | Files | + KV + runtime | Smallest fit | Source |
|---|---|---|---|---|---|
| Q8_0 ggml-org/Muse-Spark-Open-816B-A42B-GGUF | 812 GiB | 10 | unknown | over 192 GB GPU | HF files |
| BF16 ggml-org/Muse-Spark-Open-816B-A42B-GGUF | 1,528 GiB | 22 | unknown | over 192 GB GPU | HF files |
| MLX 4-bit mlx-community/Muse-Spark-Open-816B-A42B-4bit | 496 GiB | 100 | unknown | over 512 GB Mac | HF files |
No verified file: Q4_K_M, Q5_K_M, FP16, MLX 8-bit. Sizes are exact file bytes (text weights; projectors listed separately). “+” = recurrent state or other terms unknown. Smallest fit adds 0.5 GiB spare; Macs use 75% of memory.
Quality
Artificial Analysis Intelligence Index
38.9
Variant: Muse Spark Open (Xhigh) · observed 2026-10-17 · source
Rank #10 of 75 researched 2026 models.
Arena text rating
Unknown
Scores describe the evaluated hosted variant, not each quantized file. AA and Arena are never averaged.
Reported speeds 6
| Hardware | Quant | Backend | Prompt t/s | Gen t/s | Date | Source |
|---|---|---|---|---|---|---|
| Apple M3 Ultra (80-core GPU), 512 GB unified memory | 3bit (Muse-Spark-Open-816B-A42B-oQ3e) | MLX (oMLX server) | 224 | 20.6 | 2026-10-16 | omlx.ai |
| Apple M3 Ultra (80-core GPU), 512 GB unified memory | 3bit (Muse-Spark-Open-816B-A42B-oQ3e) | MLX (oMLX server) | 212 | 21.2 | 2026-10-16 | omlx.ai |
| Apple M3 Ultra (80-core GPU), 512 GB unified memory | 3bit (Muse-Spark-Open-816B-A42B-oQ3e) | MLX (oMLX server) | 227 | 20.8 | 2026-10-16 | omlx.ai |
| 2x Mac Studio M5 Ultra (512GB each), pipeline parallel, RDMA over Thunderbolt 5 (multi-machine) multi-device | NVFP4 (4-bit, ~493GB) | MLX (mlx_lm.benchmark, jaccl RDMA over Thunderbolt 5) | — | 15.7 | 2026-10-15 | machinelearning.apple.com |
| 2x Mac Studio M5 Ultra (512GB each), tensor parallel, RDMA over Thunderbolt 5 (multi-machine) multi-device | NVFP4 (4-bit, ~493GB) | MLX (mlx_lm.benchmark, jaccl RDMA over Thunderbolt 5) | — | 25 | 2026-10-15 | machinelearning.apple.com |
| 4x Mac Studio M5 Ultra (512GB each), tensor parallel, RDMA over Thunderbolt 5 (multi-machine) multi-device | NVFP4 (4-bit, ~493GB) | MLX (mlx_lm.benchmark, jaccl RDMA over Thunderbolt 5) | — | 36.7 | 2026-10-15 | machinelearning.apple.com |
Reported by the linked sources, not reproduced by LocalFit. All rows in the speed database