▸ DEVICE UNDER TEST
NVIDIA H100 NVL 94 GB — 94 GB VRAM.
▸ H100 NVL 94 GB SPEC
- BRAND
- NVIDIA
- VRAM
- 94 GB HBM3
- BANDWIDTH
- 3940 GB/s
- FP16 COMPUTE
- 241.3 TFLOPS
- FP32 COMPUTE
- 60.3 TFLOPS
- CUDA CORES
- 16,896
- TENSOR CORES
- 528
- TDP
- 400 W
- ARCHITECTURE
- Hopper
- MSRP
- $30000
▸ AI CAPABILITY
392/ 449 models @ Q4
With 94 GB VRAM and 3940 GB/s bandwidth, this GPU handles models up to 132B parameters.
Speed ≈ bandwidth / model_size × efficiency. A 7B model at Q4 runs at ~450 tok/s.
§ 01TOP MODELS FOR H100 NVL 94 GB
392 FIT · SHOWING 20| MODEL | SIZE | VRAM Q4 | TOK/S | AVG |
|---|---|---|---|---|
| DBRX 132B | 132B | 81.2 GB | 88 | 46.3 |
| Mistral Medium 3.5 | 128B | 78.7 GB | 25 | 46.6 |
| Pixtral Large 124B | 124B | 76.3 GB | 25 | 39.3 |
| Nemotron 3 Super 120B-A12B | 123.61B | 76.0 GB | 263 | 53.2 |
| Mistral-Large 123B | 123B | 75.7 GB | 26 | 33.5 |
| Devstral 2 123B | 123B | 75.7 GB | 26 | 38.1 |
| Qwen 3.5 122B A10B | 122B | 75.1 GB | 315 | 56.8 |
| Nemotron 3 Super 120B | 120B | 73.8 GB | 263 | 57.3 |
| Mistral Small 4 119B | 119B | 73.2 GB | 485 | 50.2 |
| GPT-OSS 120B | 117B | 72.0 GB | 618 | 54.1 |
| Command A 111B | 111B | 68.3 GB | 28 | 27.6 |
| GLM 4.5 Air | 110B | 67.7 GB | 263 | 51.0 |
| Qwen 1.5 110B | 110B | 67.7 GB | 29 | 33.4 |
| Llama 4 Scout 17B-16E | 109B | 67.1 GB | 185 | 33.9 |
| Cogito v2 109B MoE | 109B | 67.1 GB | 185 | — |
| Ling 2.6 Flash | 107.49B | 66.2 GB | 426 | 36.8 |
| Sarvam 105B | 105B | 64.7 GB | 30 | 48.0 |
| Command-R+ 104B | 104B | 64.1 GB | 30 | 52.7 |
| Llama-3.2-90B-Vision-Instruct | 90B | 55.5 GB | 35 | 48.5 |
| Hunyuan A13B | 80B | 49.4 GB | 242 | 81.1 |