RunsOnMy ModelsGPUs & MacsAPI pricesVRAM calculatorNew

What AI models run on the H100 SXM?

The H100 SXM has 80 GB of VRAM (3,350 GB/s). About 76.0 GB is usable for a model. 220 of 285 open models we track fit at 8K context.

220models fit (8K ctx)
218still fit at 32K ctx
201Blargest that fits
3350 GB/smemory bandwidth

Models that fit, largest first

ModelParamsBest quant (8K)Needs32K contextMax tok/s (est.)
Step 3.7 Flash201BQ2_K74.5 GBno44details
Step 3.5 Flash199BQ2_K69.0 GBno48details
Qwen3.8 Flash Next180B (62B act.)Q2_K63.9 GBQ2_K145details
Ling 3.0 flash Sante127B (5.6B act.)Q3_K_M58.3 GBQ3_K_M1239details
Ling 3.0 flash127B (5.6B act.)Q3_K_M56.0 GBQ3_K_M1292details
Ornith 1.5 122B125B (13B act.)Q3_K_M58.0 GBQ3_K_M538details
gpt oss 120b117B (5.8B act.)F1661.9 GBF161041details
gpt oss safeguard 120b117B (5.8B act.)Q8_060.1 GBQ8_01074details
GLM 4.5 Air110B (15B act.)Q3_K_M55.4 GBQ3_K_M432details
Llama 4 Scout 17B 16E Instruct109B (18B act.)Q4_K_M63.2 GBQ4_K_M309details
Llama 4 Scout 17B 16E109B (18B act.)Q4_K_M64.1 GBQ3_K_M304details
GLM 4.5V108B (14B act.)Q4_K_M61.4 GBQ4_K_M397details
GLM 4.6V108B (14B act.)Q4_K_M67.8 GBQ3_K_M358details
Hunyuan A13B Instruct80B (13B act.)Q6_K63.0 GBQ6_K314details
Kimi Dev 72B73BQ6_K63.1 GBQ5_K_M52details
Llama 3.3 70B Instruct71BQ6_K57.0 GBQ6_K58details
DeepSeek R1 Distill Llama 70B71BQ6_K57.0 GBQ6_K58details
Llama 3.1 70B71BQ6_K57.1 GBQ6_K58details
Kimi Linear 48B A3B Instruct49B (3.0B act.)Q8_049.6 GBQ8_01050details
Nous Hermes 2 Mixtral 8x7B DPO47B (13B act.)Q8_047.7 GBQ8_0245details
falcon 40b42BQ8_042.9 GBQ8_075details
flint40B (1.3B act.)Q8_040.3 GBQ8_02520details
Maxwell V1 Flash37B (2.3B act.)Q8_037.6 GBQ8_01389details
Hermes 4.3 36B36BQ8_038.5 GBQ8_087details
Seed OSS 36B Instruct36BQ8_038.5 GBQ8_087details
Qwen3.6 35B A3B36B (3.0B act.)F1667.2 GBF16565details
Ornith 1.5 35B A3B36B (3.0B act.)F1667.2 GBF16565details
Qwen3.5 35B A3B36B (3.0B act.)F1665.6 GBF16579details
Axion 1.0 35B35B (3.9B act.)F1665.6 GBF16434details
Qwen AgentWorld 35B A3B35B (3.0B act.)F1665.6 GBF16558details
Yi 34B 200K34BF1666.5 GBQ8_049details
Yi 34B34BF1666.5 GBQ8_049details
Yi 34B Chat34BF1666.5 GBQ8_049details
Yi 1.5 34B Chat34BF1666.5 GBQ8_049details
Nous Hermes 2 Yi 34B34BF1666.5 GBQ8_049details
DeepSeek R1 Distill Qwen 32B33BF1663.7 GBQ8_051details
Qwen3 32B33BF1663.7 GBQ8_051details
GLM Z1 32B 041433BF1661.8 GBF1651details
GLM 4 32B 041433BF1661.8 GBF1651details
granite 4.0 h small32B (8.8B act.)F1660.7 GBF16190details
gemma 4 31B it31BF1660.1 GBF1655details
GLM 4.7 Flash31B (4.6B act.)F1656.9 GBF16379details
Qwen3 Coder 30B A3B Instruct31B (3.0B act.)F1658.3 GBF16558details
North Mini Code 1.030B (3.3B act.)F1658.2 GBF16506details
Hy Memory 30B A3B30B (3.0B act.)F1657.4 GBF16558details
Muse Glimmer 30B30BF1652.9 GBF1660details
granite 4.2 30b29BF1657.1 GBF1657details
Qwen3.8 27B28BF1652.3 GBF1661details
Qwen3.6 27B28BF1651.5 GBF1662details
Qwen3.5 27B28BF1651.5 GBF1662details
gemma 3 27b it27BF1652.2 GBF1662details
Arete 27B27BF1651.5 GBF1662details
MiniMax H3 Image PE27BF1652.3 GBF1661details
Trinity Mini26B (2.0B act.)F1649.7 GBF16853details
GLiDE 27B26BF1650.0 GBF1664details
diffusiongemma 26B A4B it26B (4.0B act.)F1648.5 GBF16428details
gemma 4 26B A4B it26B (4.0B act.)F1648.4 GBF16428details
LFM2.5 24B A2B24B (2.0B act.)Q8_024.5 GBQ8_01575details
ERNIE 4.5 21B A3B PT22BF1641.7 GBF1677details
ERNIE 4.5 21B A3B Thinking22BF1641.7 GBF1677details
gpt oss safeguard 20b22B (4.8B act.)F1613.8 GBF161091details
gpt oss 20b21B (4.2B act.)F1613.8 GBF161211details
Groot 19B A3B 26102319B (3.0B act.)F1636.4 GBF16571details
Kimi VL A3B Thinking 250616B (3.0B act.)F1660.4 GBF16287details
Kimi VL A3B Instruct16B (3.0B act.)F1631.5 GBF16558details
Kimi VL A3B Thinking16B (3.0B act.)F1631.5 GBF16558details
deepseek moe 16b base16B (2.8B act.)F1632.8 GBF16592details
LLaDA2.1 mini16B (1.4B act.)F1631.3 GBF161169details
Moonlight 16B A3B Instruct16B (3.0B act.)F1630.7 GBF16558details
DeepSeek Coder V2 Lite Instruct16B (2.7B act.)F1630.1 GBF16629details
DeepSeek V2 Lite16B (2.7B act.)F1630.1 GBF16629details
DeepSeek V2 Lite Chat16B (2.7B act.)F1630.1 GBF16629details
Qwen2.5 Coder 14B Instruct15BF1629.7 GBF16113details
DeepSeek R1 Distill Qwen 14B15BF1629.7 GBF16113details
Qwen3 14B15BF1629.4 GBF16113details
phi 415BF1629.5 GBF16114details
gemma 3 12b it12BF1623.6 GBF16142details
Llama Guard 4 12B12BF1624.6 GBF16140details
Llama 3.2 11B Vision Instruct11BF1620.1 GBF16171details
Llama 3.2 11B Vision11BF1621.7 GBF16157details
GLM 4.6V Flash10BF1618.5 GBF16178details
GLM 4.1V 9B Thinking10BF1618.5 GBF16178details
Step3 VL 10B10BF1632.3 GBF16102details
SenseNova Loop 9B Preview9.7BF1617.8 GBF16187details
Qwen3.5 9B9.7BF1617.8 GBF16187details
Ornith 1.5 9B9.7BF1617.8 GBF16187details
Qwen3.5 9B Base9.7BF1617.8 GBF16187details
MiMo V2.6 Distill Qwen 9B9.4BF1617.8 GBF16187details
glm 4 9b9.4BF1623.2 GBF16178details
internlm3 8b instruct8.8BF1617.4 GBF16190details
granite 4.2 8b8.8BF1618.2 GBF16190details
granite 4.1 8b8.8BF1618.2 GBF16190details
Qwen3 VL 8B Instruct8.8BF1617.1 GBF16204details
Molmo2 8B8.7BF1617.1 GBF16204details
LFM2.5 8B A1B8.5B (1.0B act.)F1632.3 GBF16837details
Qwen2.5 VL 7B Instruct8.3BF1615.3 GBF16220details
UI TARS 1.5 7B8.3BF1615.3 GBF16220details
olmOCR 2 7B 10258.3BF1615.3 GBF16220details
olmOCR 7B 0225 preview8.3BF1616.6 GBF16202details
UI TARS 7B DPO8.3BF1615.3 GBF16220details
Qwen3 8B8.2BF1617.1 GBF16204details
DeepSeek R1 0528 Qwen3 8B8.2BF1617.1 GBF16204details
granite 3.3 8b instruct8.2BF1617.0 GBF16205details
granite 3.1 8b instruct8.2BF1617.0 GBF16205details
granite 3.2 8b instruct8.2BF1617.0 GBF16205details
Apertus 8B Instruct 25098.1BF1616.7 GBF16208details
Llama 3.1 8B Instruct8.0BF1616.6 GBF16208details
Llama 3.1 8B8.0BF1616.6 GBF16209details
Hermes 3 Llama 3.1 8B8.0BF1616.6 GBF16209details
DeepSeek R1 Distill Llama 8B8.0BF1631.6 GBF16104details
Meta Llama 3 8B Instruct8.0BF1616.6 GBF16209details
Meta Llama Guard 2 8B8.0BF1616.6 GBF16209details
Meta Llama 3.1 8B Instruct8.0BF1616.6 GBF16208details
Ling 3.0 tiny7.9B (1.4B act.)F1615.6 GBF161211details
MiMo 7B RL7.8BF1616.4 GBF16214details
internlm2 5 7b chat7.7BF1616.0 GBF16216details
internlm2 chat 7b7.7BF1616.0 GBF16216details
Qwen2.5 7B Instruct7.6BF1615.3 GBF16220details
Qwen2.5 Coder 7B Instruct7.6BF1615.3 GBF16220details
DeepSeek R1 Distill Qwen 7B7.6BF1615.3 GBF16220details
Falcon H1R 7B7.6BF1615.1 GBF16221details
Olmo 3 7B Instruct7.3BF1616.7 GBF16229details
Olmo 3 1025 7B7.3BF1616.7 GBF16230details
Olmo 3 7B Think7.3BF1616.7 GBF16229details
Mistral 7B v0.17.2BF1614.9 GBF16231details
Mistral 7B Instruct v0.27.2BF1614.9 GBF16231details
Mistral 7B Instruct v0.17.2BF1614.9 GBF16231details
falcon 7b7.2BF1618.4 GBF16232details
falcon 7b instruct7.2BF1618.4 GBF16232details
granite 4.0 h tiny6.9B (1.5B act.)F1613.6 GBF161142details
OLMoE 1B 7B 09246.9B (1.3B act.)F1614.4 GBF161306details
deepseek coder 7b instruct v1.56.9BF1617.2 GBF16242details
deepseek coder 6.7b instruct6.7BF1617.0 GBF16248details
deepseek coder 6.7b base6.7BF1617.0 GBF16248details
Llama 2 7b hf6.7BF1617.0 GBF16249details
Yi 6B 200K6.1BF1612.3 GBF16276details
Yi 6B6.1BF1612.3 GBF16276details
Molmo2 4B4.9BF1610.0 GBF16379details
WeVisOne VL 4B4.8BF1610.0 GBF16379details
Mage VL Cowork4.7BF1610.0 GBF16379details
Qianfan OCR4.7BF1610.1 GBF16379details
Qwen3.5 4B4.7BF169.0 GBF16398details
S5 nano4.7BF169.2 GBF16387details
LightOnOCR 3 4B4.5BF169.0 GBF16398details
SingVL S1 4B4.5BF169.0 GBF16398details
Intern Decision 4B4.5BF169.0 GBF16398details
Qwen3 VL 4B Instruct4.4BF169.3 GBF16416details
gemma 3 4b it4.3BF168.4 GBF16431details
medgemma 1.5 4b it4.3BF168.4 GBF16431details
medgemma 4b it4.3BF168.4 GBF16431details
gemma 3 4b pt4.3BF169.2 GBF16390details
Phi 3.5 vision instruct4.1BF1611.2 GBF16404details
Qwen3 4B4.0BF169.3 GBF16416details
Qwen3 4B Instruct 25074.0BF169.3 GBF16416details
granite vision 4.1 4b4.0BF167.6 GBF16492details
Phi 4 mini instruct3.8BF168.9 GBF16436details
Phi 3 mini 4k instruct3.8BF1610.6 GBF16438details
Phi 3.5 mini instruct3.8BF1610.6 GBF16438details
Phi 3 mini 128k instruct3.8BF1610.6 GBF16438details
Qwen2.5 VL 3B Instruct3.8BF166.7 GBF16542details
granite 4.2 3b3.7BF168.0 GBF16457details
granite 4.1 3b llmon preview3.7BF168.0 GBF16458details
Raon 1 nano3.5BF167.0 GBF16534details
granite 4.1 3b3.4BF167.6 GBF16492details
granite 4.0 micro3.4BF167.6 GBF16492details
DeepSeek OCR 23.4B (1.2B act.)F166.6 GBF161618details
deepseek vl2 tiny3.4B (1.2B act.)F167.4 GBF161425details
Unlimited OCR 23.3B (1.1B act.)F166.6 GBF161657details
Unlimited OCR3.3B (1.1B act.)F166.6 GBF161667details
DeepSeek OCR3.3B (1.1B act.)F166.6 GBF161667details
granite 3.1 3b a800m instruct3.3B (883M act.)F167.1 GBF161896details
Llama 3.2 3B Instruct3.2BF1613.5 GBF16260details
Llama 3.2 3B3.2BF167.5 GBF16521details
Hermes 3 Llama 3.2 3B3.2BF167.5 GBF16521details
d1 3B3.1BF1610.8 GBF16310details
LFM2.5 VL 3B3.1BF1610.8 GBF16310details
Qwen2.5 3B Instruct3.1BF166.7 GBF16543details
SmolLM3 3B3.1BF166.9 GBF16544details
SmolLM3 3B Base3.1BF166.9 GBF16545details
phi 22.8BF168.2 GBF16603details
LFM2.5 2.6B2.7BF1610.8 GBF16310details
granite 3.3 2b instruct2.5BF165.8 GBF16661details
MiniCPM5 2B2.5BF165.7 GBF16665details
WeVisOne VL 2B2.4BF165.4 GBF16822details
Qwen3.5 2B2.3BF164.5 GBF16887details
SmolVLM Instruct2.2BF165.4 GBF16924details
Qwen3 1.7B2.0BF164.8 GBF16972details
Qwen3.8 27B DSpine b162.0BF164.8 GBF16825details
d1 2B1.9BF167.2 GBF16483details
LFM3 VL 2B1.9BF167.2 GBF16483details
DeepSeek R1 Distill Qwen 1.5B1.8BF164.2 GBF16941details
LFM3 1.7B1.7BF167.2 GBF16483details
LFM3 1.7B Base1.7BF164.0 GBF16969details
SmolLM2 1.7B Instruct1.7BF165.2 GBF16978details
SmolLM2 1.7B1.7BF165.2 GBF16979details
SmolLM 1.7B1.7BF165.2 GBF16978details
granite 4.0 1b1.6BF164.3 GBF161025details
Qwen2.5 1.5B Instruct1.5BF163.8 GBF161085details
OLMo 2 0425 1B1.5BF164.4 GBF161127details
deepseek coder 1.3b instruct1.3BF164.5 GBF161244details
Llama 3.2 1B1.2BF163.2 GBF161355details
Llama 3.2 1B Instruct1.2BF165.5 GBF16676details
LFM2.5 1.2B Instruct1.2BF165.0 GBF16715details
HunyuanOCR1.1BF163.5 GBF161496details
PaddleOCR VL 1.6959MF162.5 GBF161747details
Qwen3.5 0.8B873MF162.4 GBF162209details
LightOnOCR 3 0.8B853MF162.4 GBF162209details
FastVLM 0.5B759MF162.2 GBF162207details
Qwen3 0.6B752MF162.7 GBF162796details
GOT OCR2 0716MF162.8 GBF162339details
Nemotron 3.5 Parse631MF161.9 GBF162563details
GOT OCR 2.0 hf561MF162.5 GBF162988details
SmolVLM2 500M Video Instruct507MF161.6 GBF164083details
SmolVLM 500M Instruct507MF161.6 GBF164083details
Qwen2.5 0.5B Instruct494MF161.7 GBF163390details
LFM2.5 VL 450M449MF161.9 GBF162354details
SmolLM2 360M Instruct362MF161.5 GBF164617details
SmolLM2 360M362MF161.5 GBF164617details
LFM2.5 350M354MF161.9 GBF162354details
LFM2 350M354MF161.4 GBF164725details

Questions

What is the biggest model that runs on the H100 SXM?

Of the models we track, the largest that fits at 8K context is Step 3.7 Flash (201B) at Q2_K, needing about 74.5 GB.

How much of the H100 SXM's memory can a model use?

About 76.0 GB: we assume 95% of the 80 GB is usable after driver and display overhead.

Spec source: NVIDIA specifications. Model data: Hugging Face Hub API. Speed is a bandwidth ceiling (bandwidth ÷ bytes read per token), not a benchmark. Methodology · Report an error.