NVIDIA Nemotron 3.5 Super VL 120B-A12B
Files and memory at 8K context
| Quant | File | Files | + KV + runtime | Smallest fit | Source |
|---|---|---|---|---|---|
| Q8_0 unsloth/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-GGUF | 123 GiB | 4 | 124 GiB+ | 128 GB GPU | HF files |
| MLX 4-bit Duckets/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-MLX-4bit | 64.65 GiB | 14 | 65.7 GiB+ | 96 GB Mac | HF files |
No verified file: Q4_K_M, Q5_K_M, FP16, BF16, 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
26.5
Variant: Nemotron 3.5 Super VL · observed 2026-10-17 · source
Rank #19 of 75 researched 2026 models.
Arena text rating
1,409
95% CI 1397–1421 · 2,523 votes · published 2026-10-15 · dataset
Exact listed model name; hosted reasoning/quantization may differ from local checkpoint.
Scores describe the evaluated hosted variant, not each quantized file. AA and Arena are never averaged.
Reported speeds 2
| Hardware | Quant | Backend | Prompt t/s | Gen t/s | Date | Source |
|---|---|---|---|---|---|---|
| Apple M1 Ultra (64-core GPU), 128 GB unified memory | MLX 4-bit (oMLX run, rapid-mlx style conversion) | MLX | 329 | 41.7 | 2026-10-16 | omlx.ai |
| Apple M1 Ultra (64-core GPU), 128 GB unified memory | MLX 4-bit (oMLX run, rapid-mlx style conversion) | MLX | 343 | 40.1 | 2026-10-16 | omlx.ai |
Reported by the linked sources, not reproduced by LocalFit. All rows in the speed database