H100 vs H200 for local AI
Both fit the same models we track, up to gpt-oss 120B. The H200 generates about 43% faster on any model both can run, because it has 4800 GB/s of memory bandwidth to the H100's 3350 GB/s.
Memory and bandwidth
Memory decides which models fit. Bandwidth decides how fast they generate, because every token reads the model's active weights once.
| H100 | H200 | |
|---|---|---|
| Memory | 80 GB VRAM | 141 GB VRAM |
| Bandwidth | 3350 GB/s | 4800 GB/s |
| Biggest model that fits | gpt-oss 120B | gpt-oss 120B |
| Best everyday pick | gpt-oss 120B, ~400 tok/s | gpt-oss 120B, ~580 tok/s |
Every model, side by side
Estimated generation speed at the largest memory size each was sold in (80 GB and 141 GB).
| Model | H100 | H200 |
|---|---|---|
| Qwen3 4B About 4 GB | ~740 tok/s Runs great | ~1060 tok/s Runs great |
| Qwen3 8B About 7 GB | ~350 tok/s Runs great | ~510 tok/s Runs great |
| Gemma 3 12B About 10 GB | ~230 tok/s Runs great | ~330 tok/s Runs great |
| Qwen3 14B About 11 GB | ~200 tok/s Runs great | ~280 tok/s Runs great |
| gpt-oss 20B About 16 GB | ~590 tok/s Runs great | ~840 tok/s Runs great |
| Gemma 3 27B About 19 GB | ~110 tok/s Runs great | ~160 tok/s Runs great |
| Qwen3 30B-A3B About 21 GB | ~590 tok/s Runs great | ~840 tok/s Runs great |
| Qwen3 32B About 22 GB | ~92 tok/s Runs great | ~130 tok/s Runs great |
| Llama 3.3 70B About 45 GB | ~43 tok/s Runs great | ~61 tok/s Runs great |
| gpt-oss 120B About 67 GB | ~400 tok/s Runs great | ~580 tok/s Runs great |
| Qwen3 235B-A22B About 144 GB | Needs 144 GB | Needs 144 GB |
| DeepSeek R1 671B About 406 GB | Needs 406 GB | Needs 406 GB |
The H200, for speed. They fit the same models. Price matters as much as either number: check used prices on the builds page, where every parts list has a measured or estimated speed.
Both columns use the same formula: memory bandwidth divided by the weights read per token, calibrated on our own rack, with Ollama's default 4-bit sizes. Real results are usually within about a third. Measured either one? Send us the number.