Coding AIMIT
NVIDIA Nemotron 3 Super
nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16
Parameters
120B
Context
1024K
Downloads
818.9K
Likes
415
Architecture
moe_transformer
GPU Requirements
VRAM / RAM / disk estimates by quantization and usage scenario
| Quantization | Minimum | Recommended | Production |
|---|---|---|---|
| FP1616bit | 287.7 GB RAM 431.6 · Disk 387.7 | 316.3 GB RAM 474.5 · Disk 413.7 | 426.1 GB RAM 639.1 · Disk 452.7 |
| FP88bit | 146.3 GB RAM 219.5 · Disk 213.3 | 168.5 GB RAM 252.8 · Disk 239.3 | 271.8 GB RAM 407.7 · Disk 278.3 |
| INT88bit | 146.3 GB RAM 219.5 · Disk 213.3 | 168.5 GB RAM 252.8 · Disk 239.3 | 271.8 GB RAM 407.7 · Disk 278.3 |
| AWQ4bit | 78.3 GB RAM 117.4 · Disk 129.4 | 97.4 GB RAM 146.1 · Disk 155.4 | 197.6 GB RAM 296.4 · Disk 194.4 |
| GPTQ4bit | 79.2 GB RAM 118.8 · Disk 130.5 | 98.3 GB RAM 147.4 · Disk 156.5 | 198.6 GB RAM 297.9 · Disk 195.5 |
| GGUF4bit | 80.9 GB RAM 121.4 · Disk 132.7 | 100.1 GB RAM 150.2 · Disk 158.7 | 200.5 GB RAM 300.8 · Disk 197.7 |
GPU Recommendations
Best ValueBEST
4× H200 SXM
564 GB VRAM
$10,220/mo
≈ 198.7 tok/s
Effective production throughput ≈ 139.1 tok/s
score 0.3154
Cheapest
8× Tesla V100
128 GB VRAM
$162/mo
≈ 0 tok/s (N/A — benchmark unavailable)
Effective production throughput ≈ 0 tok/s
score 0.5538
Performance
8× H200 SXM
1128 GB VRAM
$20,440/mo
≈ 265 tok/s
Effective production throughput ≈ 185.5 tok/s
score 0.2197
Min Complexity
1× H200 SXM
141 GB VRAM
$2,555/mo
≈ 82.8 tok/s
Effective production throughput ≈ 58 tok/s
score 0.6552
💬 Community — real deployment experience
0 Articles · 0 Benchmarks
GPUs that fit (single card)
H200 SXM
141 GB VRAM
$2,555
/mo · lambda
B200
180 GB VRAM
$3,013
/mo · vast
B200 SXM
180 GB VRAM
—