RunsOnMy ModelsGPUs & MacsAPI pricesVRAM calculatorNew

What AI models run on the H100 PCIe?

The H100 PCIe has 80 GB of VRAM (2,000 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
2000 GB/smemory bandwidth

Models that fit, largest first

ModelParamsBest quant (8K)Needs32K contextMax tok/s (est.)
Step 3.7 Flash201BQ2_K74.5 GBno26details
Step 3.5 Flash199BQ2_K69.0 GBno29details
Qwen3.8 Flash Next180B (62B act.)Q2_K63.9 GBQ2_K87details
Ling 3.0 flash Sante127B (5.6B act.)Q3_K_M58.3 GBQ3_K_M740details
Ling 3.0 flash127B (5.6B act.)Q3_K_M56.0 GBQ3_K_M771details
Ornith 1.5 122B125B (13B act.)Q3_K_M58.0 GBQ3_K_M321details
gpt oss 120b117B (5.8B act.)F1661.9 GBF16622details
gpt oss safeguard 120b117B (5.8B act.)Q8_060.1 GBQ8_0641details
GLM 4.5 Air110B (15B act.)Q3_K_M55.4 GBQ3_K_M258details
Llama 4 Scout 17B 16E Instruct109B (18B act.)Q4_K_M63.2 GBQ4_K_M184details
Llama 4 Scout 17B 16E109B (18B act.)Q4_K_M64.1 GBQ3_K_M181details
GLM 4.5V108B (14B act.)Q4_K_M61.4 GBQ4_K_M237details
GLM 4.6V108B (14B act.)Q4_K_M67.8 GBQ3_K_M214details
Hunyuan A13B Instruct80B (13B act.)Q6_K63.0 GBQ6_K188details
Kimi Dev 72B73BQ6_K63.1 GBQ5_K_M31details
Llama 3.3 70B Instruct71BQ6_K57.0 GBQ6_K35details
DeepSeek R1 Distill Llama 70B71BQ6_K57.0 GBQ6_K35details
Llama 3.1 70B71BQ6_K57.1 GBQ6_K35details
Kimi Linear 48B A3B Instruct49B (3.0B act.)Q8_049.6 GBQ8_0627details
Nous Hermes 2 Mixtral 8x7B DPO47B (13B act.)Q8_047.7 GBQ8_0146details
falcon 40b42BQ8_042.9 GBQ8_045details
flint40B (1.3B act.)Q8_040.3 GBQ8_01505details
Maxwell V1 Flash37B (2.3B act.)Q8_037.6 GBQ8_0829details
Hermes 4.3 36B36BQ8_038.5 GBQ8_052details
Seed OSS 36B Instruct36BQ8_038.5 GBQ8_052details
Qwen3.6 35B A3B36B (3.0B act.)F1667.2 GBF16337details
Ornith 1.5 35B A3B36B (3.0B act.)F1667.2 GBF16337details
Qwen3.5 35B A3B36B (3.0B act.)F1665.6 GBF16345details
Axion 1.0 35B35B (3.9B act.)F1665.6 GBF16259details
Qwen AgentWorld 35B A3B35B (3.0B act.)F1665.6 GBF16333details
Yi 34B 200K34BF1666.5 GBQ8_029details
Yi 34B34BF1666.5 GBQ8_029details
Yi 34B Chat34BF1666.5 GBQ8_029details
Yi 1.5 34B Chat34BF1666.5 GBQ8_029details
Nous Hermes 2 Yi 34B34BF1666.5 GBQ8_029details
DeepSeek R1 Distill Qwen 32B33BF1663.7 GBQ8_031details
Qwen3 32B33BF1663.7 GBQ8_031details
GLM Z1 32B 041433BF1661.8 GBF1631details
GLM 4 32B 041433BF1661.8 GBF1631details
granite 4.0 h small32B (8.8B act.)F1660.7 GBF16114details
gemma 4 31B it31BF1660.1 GBF1633details
GLM 4.7 Flash31B (4.6B act.)F1656.9 GBF16226details
Qwen3 Coder 30B A3B Instruct31B (3.0B act.)F1658.3 GBF16333details
North Mini Code 1.030B (3.3B act.)F1658.2 GBF16302details
Hy Memory 30B A3B30B (3.0B act.)F1657.4 GBF16333details
Muse Glimmer 30B30BF1652.9 GBF1636details
granite 4.2 30b29BF1657.1 GBF1634details
Qwen3.8 27B28BF1652.3 GBF1637details
Qwen3.6 27B28BF1651.5 GBF1637details
Qwen3.5 27B28BF1651.5 GBF1637details
gemma 3 27b it27BF1652.2 GBF1637details
Arete 27B27BF1651.5 GBF1637details
MiniMax H3 Image PE27BF1652.3 GBF1637details
Trinity Mini26B (2.0B act.)F1649.7 GBF16509details
GLiDE 27B26BF1650.0 GBF1638details
diffusiongemma 26B A4B it26B (4.0B act.)F1648.5 GBF16255details
gemma 4 26B A4B it26B (4.0B act.)F1648.4 GBF16255details
LFM2.5 24B A2B24B (2.0B act.)Q8_024.5 GBQ8_0940details
ERNIE 4.5 21B A3B PT22BF1641.7 GBF1646details
ERNIE 4.5 21B A3B Thinking22BF1641.7 GBF1646details
gpt oss safeguard 20b22B (4.8B act.)F1613.8 GBF16651details
gpt oss 20b21B (4.2B act.)F1613.8 GBF16723details
Groot 19B A3B 26102319B (3.0B act.)F1636.4 GBF16341details
Kimi VL A3B Thinking 250616B (3.0B act.)F1660.4 GBF16171details
Kimi VL A3B Instruct16B (3.0B act.)F1631.5 GBF16333details
Kimi VL A3B Thinking16B (3.0B act.)F1631.5 GBF16333details
deepseek moe 16b base16B (2.8B act.)F1632.8 GBF16354details
LLaDA2.1 mini16B (1.4B act.)F1631.3 GBF16698details
Moonlight 16B A3B Instruct16B (3.0B act.)F1630.7 GBF16333details
DeepSeek Coder V2 Lite Instruct16B (2.7B act.)F1630.1 GBF16376details
DeepSeek V2 Lite16B (2.7B act.)F1630.1 GBF16376details
DeepSeek V2 Lite Chat16B (2.7B act.)F1630.1 GBF16376details
Qwen2.5 Coder 14B Instruct15BF1629.7 GBF1668details
DeepSeek R1 Distill Qwen 14B15BF1629.7 GBF1668details
Qwen3 14B15BF1629.4 GBF1668details
phi 415BF1629.5 GBF1668details
gemma 3 12b it12BF1623.6 GBF1685details
Llama Guard 4 12B12BF1624.6 GBF1683details
Llama 3.2 11B Vision Instruct11BF1620.1 GBF16102details
Llama 3.2 11B Vision11BF1621.7 GBF1694details
GLM 4.6V Flash10BF1618.5 GBF16106details
GLM 4.1V 9B Thinking10BF1618.5 GBF16106details
Step3 VL 10B10BF1632.3 GBF1661details
SenseNova Loop 9B Preview9.7BF1617.8 GBF16112details
Qwen3.5 9B9.7BF1617.8 GBF16112details
Ornith 1.5 9B9.7BF1617.8 GBF16112details
Qwen3.5 9B Base9.7BF1617.8 GBF16112details
MiMo V2.6 Distill Qwen 9B9.4BF1617.8 GBF16112details
glm 4 9b9.4BF1623.2 GBF16106details
internlm3 8b instruct8.8BF1617.4 GBF16114details
granite 4.2 8b8.8BF1618.2 GBF16114details
granite 4.1 8b8.8BF1618.2 GBF16114details
Qwen3 VL 8B Instruct8.8BF1617.1 GBF16122details
Molmo2 8B8.7BF1617.1 GBF16122details
LFM2.5 8B A1B8.5B (1.0B act.)F1632.3 GBF16500details
Qwen2.5 VL 7B Instruct8.3BF1615.3 GBF16131details
UI TARS 1.5 7B8.3BF1615.3 GBF16131details
olmOCR 2 7B 10258.3BF1615.3 GBF16131details
olmOCR 7B 0225 preview8.3BF1616.6 GBF16121details
UI TARS 7B DPO8.3BF1615.3 GBF16131details
Qwen3 8B8.2BF1617.1 GBF16122details
DeepSeek R1 0528 Qwen3 8B8.2BF1617.1 GBF16122details
granite 3.3 8b instruct8.2BF1617.0 GBF16122details
granite 3.1 8b instruct8.2BF1617.0 GBF16122details
granite 3.2 8b instruct8.2BF1617.0 GBF16122details
Apertus 8B Instruct 25098.1BF1616.7 GBF16124details
Llama 3.1 8B Instruct8.0BF1616.6 GBF16124details
Llama 3.1 8B8.0BF1616.6 GBF16125details
Hermes 3 Llama 3.1 8B8.0BF1616.6 GBF16125details
DeepSeek R1 Distill Llama 8B8.0BF1631.6 GBF1662details
Meta Llama 3 8B Instruct8.0BF1616.6 GBF16125details
Meta Llama Guard 2 8B8.0BF1616.6 GBF16125details
Meta Llama 3.1 8B Instruct8.0BF1616.6 GBF16124details
Ling 3.0 tiny7.9B (1.4B act.)F1615.6 GBF16723details
MiMo 7B RL7.8BF1616.4 GBF16128details
internlm2 5 7b chat7.7BF1616.0 GBF16129details
internlm2 chat 7b7.7BF1616.0 GBF16129details
Qwen2.5 7B Instruct7.6BF1615.3 GBF16131details
Qwen2.5 Coder 7B Instruct7.6BF1615.3 GBF16131details
DeepSeek R1 Distill Qwen 7B7.6BF1615.3 GBF16131details
Falcon H1R 7B7.6BF1615.1 GBF16132details
Olmo 3 7B Instruct7.3BF1616.7 GBF16137details
Olmo 3 1025 7B7.3BF1616.7 GBF16137details
Olmo 3 7B Think7.3BF1616.7 GBF16137details
Mistral 7B v0.17.2BF1614.9 GBF16138details
Mistral 7B Instruct v0.27.2BF1614.9 GBF16138details
Mistral 7B Instruct v0.17.2BF1614.9 GBF16138details
falcon 7b7.2BF1618.4 GBF16139details
falcon 7b instruct7.2BF1618.4 GBF16139details
granite 4.0 h tiny6.9B (1.5B act.)F1613.6 GBF16682details
OLMoE 1B 7B 09246.9B (1.3B act.)F1614.4 GBF16780details
deepseek coder 7b instruct v1.56.9BF1617.2 GBF16145details
deepseek coder 6.7b instruct6.7BF1617.0 GBF16148details
deepseek coder 6.7b base6.7BF1617.0 GBF16148details
Llama 2 7b hf6.7BF1617.0 GBF16148details
Yi 6B 200K6.1BF1612.3 GBF16165details
Yi 6B6.1BF1612.3 GBF16165details
Molmo2 4B4.9BF1610.0 GBF16226details
WeVisOne VL 4B4.8BF1610.0 GBF16226details
Mage VL Cowork4.7BF1610.0 GBF16227details
Qianfan OCR4.7BF1610.1 GBF16226details
Qwen3.5 4B4.7BF169.0 GBF16237details
S5 nano4.7BF169.2 GBF16231details
LightOnOCR 3 4B4.5BF169.0 GBF16237details
SingVL S1 4B4.5BF169.0 GBF16237details
Intern Decision 4B4.5BF169.0 GBF16237details
Qwen3 VL 4B Instruct4.4BF169.3 GBF16248details
gemma 3 4b it4.3BF168.4 GBF16257details
medgemma 1.5 4b it4.3BF168.4 GBF16257details
medgemma 4b it4.3BF168.4 GBF16257details
gemma 3 4b pt4.3BF169.2 GBF16233details
Phi 3.5 vision instruct4.1BF1611.2 GBF16241details
Qwen3 4B4.0BF169.3 GBF16249details
Qwen3 4B Instruct 25074.0BF169.3 GBF16248details
granite vision 4.1 4b4.0BF167.6 GBF16294details
Phi 4 mini instruct3.8BF168.9 GBF16260details
Phi 3 mini 4k instruct3.8BF1610.6 GBF16262details
Phi 3.5 mini instruct3.8BF1610.6 GBF16262details
Phi 3 mini 128k instruct3.8BF1610.6 GBF16262details
Qwen2.5 VL 3B Instruct3.8BF166.7 GBF16324details
granite 4.2 3b3.7BF168.0 GBF16273details
granite 4.1 3b llmon preview3.7BF168.0 GBF16273details
Raon 1 nano3.5BF167.0 GBF16319details
granite 4.1 3b3.4BF167.6 GBF16294details
granite 4.0 micro3.4BF167.6 GBF16294details
DeepSeek OCR 23.4B (1.2B act.)F166.6 GBF16966details
deepseek vl2 tiny3.4B (1.2B act.)F167.4 GBF16851details
Unlimited OCR 23.3B (1.1B act.)F166.6 GBF16989details
Unlimited OCR3.3B (1.1B act.)F166.6 GBF16995details
DeepSeek OCR3.3B (1.1B act.)F166.6 GBF16995details
granite 3.1 3b a800m instruct3.3B (883M act.)F167.1 GBF161132details
Llama 3.2 3B Instruct3.2BF1613.5 GBF16155details
Llama 3.2 3B3.2BF167.5 GBF16311details
Hermes 3 Llama 3.2 3B3.2BF167.5 GBF16311details
d1 3B3.1BF1610.8 GBF16185details
LFM2.5 VL 3B3.1BF1610.8 GBF16185details
Qwen2.5 3B Instruct3.1BF166.7 GBF16324details
SmolLM3 3B3.1BF166.9 GBF16325details
SmolLM3 3B Base3.1BF166.9 GBF16325details
phi 22.8BF168.2 GBF16360details
LFM2.5 2.6B2.7BF1610.8 GBF16185details
granite 3.3 2b instruct2.5BF165.8 GBF16395details
MiniCPM5 2B2.5BF165.7 GBF16397details
WeVisOne VL 2B2.4BF165.4 GBF16491details
Qwen3.5 2B2.3BF164.5 GBF16530details
SmolVLM Instruct2.2BF165.4 GBF16552details
Qwen3 1.7B2.0BF164.8 GBF16580details
Qwen3.8 27B DSpine b162.0BF164.8 GBF16492details
d1 2B1.9BF167.2 GBF16288details
LFM3 VL 2B1.9BF167.2 GBF16288details
DeepSeek R1 Distill Qwen 1.5B1.8BF164.2 GBF16562details
LFM3 1.7B1.7BF167.2 GBF16288details
LFM3 1.7B Base1.7BF164.0 GBF16578details
SmolLM2 1.7B Instruct1.7BF165.2 GBF16584details
SmolLM2 1.7B1.7BF165.2 GBF16584details
SmolLM 1.7B1.7BF165.2 GBF16584details
granite 4.0 1b1.6BF164.3 GBF16612details
Qwen2.5 1.5B Instruct1.5BF163.8 GBF16648details
OLMo 2 0425 1B1.5BF164.4 GBF16673details
deepseek coder 1.3b instruct1.3BF164.5 GBF16743details
Llama 3.2 1B1.2BF163.2 GBF16809details
Llama 3.2 1B Instruct1.2BF165.5 GBF16403details
LFM2.5 1.2B Instruct1.2BF165.0 GBF16427details
HunyuanOCR1.1BF163.5 GBF16893details
PaddleOCR VL 1.6959MF162.5 GBF161043details
Qwen3.5 0.8B873MF162.4 GBF161319details
LightOnOCR 3 0.8B853MF162.4 GBF161319details
FastVLM 0.5B759MF162.2 GBF161318details
Qwen3 0.6B752MF162.7 GBF161669details
GOT OCR2 0716MF162.8 GBF161397details
Nemotron 3.5 Parse631MF161.9 GBF161530details
GOT OCR 2.0 hf561MF162.5 GBF161784details
SmolVLM2 500M Video Instruct507MF161.6 GBF162438details
SmolVLM 500M Instruct507MF161.6 GBF162438details
Qwen2.5 0.5B Instruct494MF161.7 GBF162024details
LFM2.5 VL 450M449MF161.9 GBF161406details
SmolLM2 360M Instruct362MF161.5 GBF162757details
SmolLM2 360M362MF161.5 GBF162757details
LFM2.5 350M354MF161.9 GBF161406details
LFM2 350M354MF161.4 GBF162821details

Questions

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

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 PCIe'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.