What can the H100 SXM run?
| Text models: 110 | Visual models: 9 | Strongest first |
- Memory
- 80 GB
- Bandwidth
- 3,350 GB/s
- Maker
- NVIDIA
Text models that run on it
| Model | Best fit | Context | Speed | Strongest at |
|---|---|---|---|---|
| Qwen3.8 Flash-Next | 2-bit | 256K | 257 tok/s | Chat, Coding, Agents |
| Qwen3.8 27B | 8-bit | 256K | 71 tok/s | Chat, Coding, Agents, Docs |
| Muse Glimmer 30B | 8-bit | 128K | 69 tok/s | Docs |
| Mistral Medium 3.5 128B | 3-bit | 48K | 35 tok/s | Chat, Coding, Agents, Docs |
| Qwen3.5 122B A10B | 4-bit | 176K | 181 tok/s | Chat, Coding, Agents, Docs |
| Nemotron 3 Super 120B A12B | 4-bit | 256K | 118 tok/s | Chat, Coding, Agents, Docs |
| Nemotron 3.5 Bolt 74B A9B | 8-bit | 112K | 106 tok/s | Chat, Coding, Agents, Docs |
| Qwen3.6 35B A3B | 8-bit | 256K | 264 tok/s | Chat, Coding, Agents, Docs |
| Gemma 4 31B | 8-bit | 256K | 63 tok/s | Chat, Coding, Agents, Docs |
| Gemma 4 26B A4B | 8-bit | 256K | 284 tok/s | Chat, Coding, Agents, Docs |
| GLM-4.7 Flash | 8-bit | 198K | 239 tok/s | Chat, Coding, Agents |
| Mistral Small 4 119B | 4-bit | 256K | 260 tok/s | Chat, Docs |
| gpt-oss 120B | MXFP4 | 128K | 269 tok/s | Chat, Coding |
| Qwen3 Coder Next 80B | 6-bit | 256K | 252 tok/s | Coding, Agents |
| Nemotron 3.5 Lightning 30B A3B | 8-bit | 256K | 222 tok/s | Agents, Docs |
| Devstral Small 2 24B | FP8 | 320K | 84 tok/s | Coding |
| LFM2.5 24B A2B | 8-bit | 128K | 302 tok/s | Agents |
| Mistral Small 3.2 24B | 8-bit | 128K | 85 tok/s | Chat, Coding, Agents |
| Llama 3.3 70B | 8-bit | 16K | 30 tok/s | Chat, Coding, Agents, Docs |
| Olmo 3.1 32B | 8-bit | 64K | 61 tok/s | Chat, Coding, Agents, Docs |
| Qwen3 14B | 8-bit | 40K | 133 tok/s | Not rated |
| Qwen3 32B | 8-bit | 40K | 61 tok/s | Not rated |
| Qwen3.6 27B | 8-bit | 256K | 71 tok/s | Text and images |
| Qwen2.5 Coder 14B Instruct | 8-bit | 32K | 129 tok/s | Coding |
| Qwen3 Coder 30B A3B Instruct | 8-bit | 256K | 236 tok/s | Coding |
| Qwen3.5 27B | 8-bit | 256K | 71 tok/s | Text and images |
| Qwen3.5 35B A3B | 8-bit | 256K | 264 tok/s | Text and images |
| NVIDIA Nemotron 3 Nano 30B A3B BF16 | 8-bit | 256K | 222 tok/s | Not rated |
| GLM 4.5 Air | 4-bit | 64K | 151 tok/s | Not rated |
| Qwen3 30B A3B | 8-bit | 40K | 236 tok/s | Not rated |
| Qwen3 30B A3B Instruct 2507 | 8-bit | 256K | 236 tok/s | Not rated |
| Laguna XS.2 | 8-bit | 256K | 265 tok/s | Not rated |
| OTel 2.0 LLM 31B IT | 8-bit | 256K | 63 tok/s | Not rated |
| Qwen2.5 32B Instruct | 8-bit | 32K | 61 tok/s | Not rated |
| MiniCPM V 4.7 35B A3B | 8-bit | 256K | 264 tok/s | Text and images |
| Qwen2.5 14B Instruct | 8-bit | 32K | 129 tok/s | Not rated |
| Laguna XS 2.1 | 8-bit | 256K | 265 tok/s | Not rated |
| Diffusiongemma 26B A4B IT | 8-bit | 256K | 277 tok/s | Text and images |
| Qwen2.5 VL 32B Instruct | 8-bit | 125K | 60 tok/s | Text and images |
| Qwen2.5 Coder 32B Instruct | 8-bit | 32K | 61 tok/s | Coding |
| Laguna S 2.1 | 4-bit | 184K | 197 tok/s | Not rated |
| Qwen3 VL 30B A3B Instruct | 8-bit | 256K | 236 tok/s | Text and images |
| DeepSeek R1 Distill Qwen 32B | 8-bit | 128K | 61 tok/s | Not rated |
| Gemma 3 27B IT | 8-bit | 128K | 73 tok/s | Text and images |
| Ornith 1.5 35B A3B | 8-bit | 256K | 264 tok/s | Not rated |
| Ornith 1.5 122B | 4-bit | 176K | 160 tok/s | Not rated |
| GPT Neox 20B | 8-bit | 2K | 95 tok/s | Not rated |
| Qwen3 VL 32B Instruct | 8-bit | 168K | 60 tok/s | Text and images |
| DeepSeek R1 Distill Qwen 14B | 8-bit | 128K | 129 tok/s | Not rated |
| Qwen2.5 14B Instruct 1M | 8-bit | 320K | 129 tok/s | Not rated |
| Qwen2.5 72B Instruct | 6-bit | 32K | 37 tok/s | Not rated |
| Qwen3 Next 80B A3B Instruct | 6-bit | 256K | 252 tok/s | Not rated |
| Llama 4 Scout 17B 16E Instruct | 5-bit | 16K | 120 tok/s | Text and images |
| Qwen2.5 VL 72B Instruct | 6-bit | 60K | 36 tok/s | Text and images |
| Qwen1.5 7B | 8-bit | 32K | 239 tok/s | Not rated |
| Solar Open 100B | 5-bit | 36K | 153 tok/s | Not rated |
| Qwen3 30B A3B Thinking 2507 | 8-bit | 256K | 236 tok/s | Not rated |
| NVIDIA Nemotron 3.5 Super VL 120B A12B BF16 | 4-bit | 576K | 118 tok/s | Text and images |
| Gnani Evon V3.3 30B A3B | 8-bit | 128K | 222 tok/s | Not rated |
| EXAONE 3.5 32B Instruct | 8-bit | 32K | 62 tok/s | Not rated |
Best fit is the highest quality, up to 8-bit, that leaves the use case’s context. Ranked by our use-case scores, then by Hugging Face downloads. Speeds are rough, plus or minus 25%. The top 60 of 110 are shown.
Image and video models
| Model | Best fit | Per output | Strongest at |
|---|---|---|---|
| MiniMax H3 | 16-bit | 78 s | Video |
| LTX-2.5 | 16-bit | 26 s | Video |
| Qwen-Image | 16-bit | 36 s | Images |
| FLUX.2 [dev] | 16-bit | 16 s | Images |
| Wan2.2 I2V A14B | 16-bit | 5.4 min | Video |
| Wan2.1 I2V 14B | 16-bit | 5.4 min | Video |
| Wan 2.1 T2V-14B | 16-bit | 6.6 min | Video |
| Qwen Image 2512 | 16-bit | 36 s | Images |
| Wan 2.2 A14B | 16-bit | 5.4 min | Video |
Only just too big
- MiniMax M3
- Qwen3.5 397B A17B
- MiniMax M2.7
- Inkling Small
- MiniMax M2.5
- Laguna M.1
- Qwen3 VL 235B A22B Instruct
- Qwen3 235B A22B
- Ornith 1.5 397B
- Step 3.5 Flash
- MiniMax M2
- Qwen3 VL 235B A22B Thinking
- Qwen3 235B A22B Instruct 2507
- Qwen3 Coder 480B A35B Instruct
- DeepSeek V2 Chat
- Command A Plus 05 2026 Bf16
- MiniMax M2.1
- Beam 1.0 501B A23B
- Intern S1
- Step 3.7 Flash
These need up to a quarter more memory than the H100 SXM has. Spilling into system RAM or a lower precision can get them running, slowly.
Questions
What is the best LLM for the H100 SXM?
Of the open models we track, the strongest that run well on the H100 SXM are Qwen3.8 Flash-Next, Qwen3.8 27B, Muse Glimmer 30B. 110 text models fit it in all.
Calculated from published specs, not benchmarked. All machines