NVIDIA B100 192GB vs NVIDIA H200 for local LLMs
0 reports on the NVIDIA B100 192GB and 2 on the NVIDIA H200
Which is faster for local LLMs?
Nobody has reported the same model at the same quant on both cards yet, so there's no measured head-to-head. Memory bandwidth, which caps how fast a card can write, is 1.7x higher on the NVIDIA B100 192GB (8,000 GB/s on the NVIDIA B100 192GB, 4,800 on the NVIDIA H200). The estimates below compare them on the models people run most.
NVIDIA B100 192GB
- Memory
- 192GB
- Memory bandwidth
- 8,000 GB/s
- FP16 compute
- 1750 TFLOPS
- Reports
- 0
NVIDIA H200
- Memory
- 141GB
- Memory bandwidth
- 4,800 GB/s
- FP16 compute
- 989 TFLOPS
- Reports
- 2
Estimated by the calculator
The quant the calculator recommends for each card and its estimated speed at a 32,768-token context with 32 GB of system RAM, from memory bandwidth and the model's shape.
| Model | NVIDIA B100 192GB | NVIDIA H200 |
|---|---|---|
| Qwen3.8 27B | 81.5 t/sFP16 | 48.9 t/sFP16 |
| Qwen3.8 125B · 6B active | 278.7 t/sQ8_0 | 191.1 t/sQ6_K |
| DeepSeek V4 Flash 284B · 13B active | 367.2 t/sQ4_K_M | 362.5 t/sQ3_K_S |
| Qwen3.6 35B · 3B active | 293.3 t/sFP16 | 176.0 t/sFP16 |
| Qwen3.6 27B | 81.5 t/sFP16 | 48.9 t/sFP16 |
| Gemma 4 26B · 4B active | 220.0 t/sFP16 | 132.0 t/sFP16 |
Change the context, RAM or model in the calculator: NVIDIA B100 192GB · NVIDIA H200
Every report on each card, with its source: NVIDIA B100 192GB · NVIDIA H200