Skip to content

What can the H100 SXM run?

What fits on the H100 SXM
Text models: 110Visual models: 9Strongest first
Memory
80 GB
Bandwidth
3,350 GB/s
Maker
NVIDIA

Open the calculator on this machine

Text models that run on it

Text models that run on the H100 SXM, best first
ModelBest fitContextSpeedStrongest at
Qwen3.8 Flash-Next2-bit256K257 tok/sChat, Coding, Agents
Qwen3.8 27B8-bit256K71 tok/sChat, Coding, Agents, Docs
Muse Glimmer 30B8-bit128K69 tok/sDocs
Mistral Medium 3.5 128B3-bit48K35 tok/sChat, Coding, Agents, Docs
Qwen3.5 122B A10B4-bit176K181 tok/sChat, Coding, Agents, Docs
Nemotron 3 Super 120B A12B4-bit256K118 tok/sChat, Coding, Agents, Docs
Nemotron 3.5 Bolt 74B A9B8-bit112K106 tok/sChat, Coding, Agents, Docs
Qwen3.6 35B A3B8-bit256K264 tok/sChat, Coding, Agents, Docs
Gemma 4 31B8-bit256K63 tok/sChat, Coding, Agents, Docs
Gemma 4 26B A4B8-bit256K284 tok/sChat, Coding, Agents, Docs
GLM-4.7 Flash8-bit198K239 tok/sChat, Coding, Agents
Mistral Small 4 119B4-bit256K260 tok/sChat, Docs
gpt-oss 120BMXFP4128K269 tok/sChat, Coding
Qwen3 Coder Next 80B6-bit256K252 tok/sCoding, Agents
Nemotron 3.5 Lightning 30B A3B8-bit256K222 tok/sAgents, Docs
Devstral Small 2 24BFP8320K84 tok/sCoding
LFM2.5 24B A2B8-bit128K302 tok/sAgents
Mistral Small 3.2 24B8-bit128K85 tok/sChat, Coding, Agents
Llama 3.3 70B8-bit16K30 tok/sChat, Coding, Agents, Docs
Olmo 3.1 32B8-bit64K61 tok/sChat, Coding, Agents, Docs
Qwen3 14B8-bit40K133 tok/sNot rated
Qwen3 32B8-bit40K61 tok/sNot rated
Qwen3.6 27B8-bit256K71 tok/sText and images
Qwen2.5 Coder 14B Instruct8-bit32K129 tok/sCoding
Qwen3 Coder 30B A3B Instruct8-bit256K236 tok/sCoding
Qwen3.5 27B8-bit256K71 tok/sText and images
Qwen3.5 35B A3B8-bit256K264 tok/sText and images
NVIDIA Nemotron 3 Nano 30B A3B BF168-bit256K222 tok/sNot rated
GLM 4.5 Air4-bit64K151 tok/sNot rated
Qwen3 30B A3B8-bit40K236 tok/sNot rated
Qwen3 30B A3B Instruct 25078-bit256K236 tok/sNot rated
Laguna XS.28-bit256K265 tok/sNot rated
OTel 2.0 LLM 31B IT8-bit256K63 tok/sNot rated
Qwen2.5 32B Instruct8-bit32K61 tok/sNot rated
MiniCPM V 4.7 35B A3B8-bit256K264 tok/sText and images
Qwen2.5 14B Instruct8-bit32K129 tok/sNot rated
Laguna XS 2.18-bit256K265 tok/sNot rated
Diffusiongemma 26B A4B IT8-bit256K277 tok/sText and images
Qwen2.5 VL 32B Instruct8-bit125K60 tok/sText and images
Qwen2.5 Coder 32B Instruct8-bit32K61 tok/sCoding
Laguna S 2.14-bit184K197 tok/sNot rated
Qwen3 VL 30B A3B Instruct8-bit256K236 tok/sText and images
DeepSeek R1 Distill Qwen 32B8-bit128K61 tok/sNot rated
Gemma 3 27B IT8-bit128K73 tok/sText and images
Ornith 1.5 35B A3B8-bit256K264 tok/sNot rated
Ornith 1.5 122B4-bit176K160 tok/sNot rated
GPT Neox 20B8-bit2K95 tok/sNot rated
Qwen3 VL 32B Instruct8-bit168K60 tok/sText and images
DeepSeek R1 Distill Qwen 14B8-bit128K129 tok/sNot rated
Qwen2.5 14B Instruct 1M8-bit320K129 tok/sNot rated
Qwen2.5 72B Instruct6-bit32K37 tok/sNot rated
Qwen3 Next 80B A3B Instruct6-bit256K252 tok/sNot rated
Llama 4 Scout 17B 16E Instruct5-bit16K120 tok/sText and images
Qwen2.5 VL 72B Instruct6-bit60K36 tok/sText and images
Qwen1.5 7B8-bit32K239 tok/sNot rated
Solar Open 100B5-bit36K153 tok/sNot rated
Qwen3 30B A3B Thinking 25078-bit256K236 tok/sNot rated
NVIDIA Nemotron 3.5 Super VL 120B A12B BF164-bit576K118 tok/sText and images
Gnani Evon V3.3 30B A3B8-bit128K222 tok/sNot rated
EXAONE 3.5 32B Instruct8-bit32K62 tok/sNot 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

Image and video models that run on the H100 SXM
ModelBest fitPer outputStrongest at
MiniMax H316-bit78 sVideo
LTX-2.516-bit26 sVideo
Qwen-Image16-bit36 sImages
FLUX.2 [dev]16-bit16 sImages
Wan2.2 I2V A14B16-bit5.4 minVideo
Wan2.1 I2V 14B16-bit5.4 minVideo
Wan 2.1 T2V-14B16-bit6.6 minVideo
Qwen Image 251216-bit36 sImages
Wan 2.2 A14B16-bit5.4 minVideo

Only just too big

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