RunsOnMy ModelsGPUs & MacsAPI pricesVRAM calculatorNew

What AI models run on the RTX 4080?

The RTX 4080 has 16 GB of VRAM (717 GB/s). About 15.2 GB is usable for a model. 194 of 285 open models we track fit at 8K context.

194models fit (8K ctx)
180still fit at 32K ctx
40Blargest that fits
717 GB/smemory bandwidth

Models that fit, largest first

ModelParamsBest quant (8K)Needs32K contextMax tok/s (est.)
flint40B (1.3B act.)Q2_K14.8 GBno1529details
Maxwell V1 Flash37B (2.3B act.)Q2_K14.0 GBQ2_K842details
Qwen3.6 35B A3B36B (3.0B act.)Q2_K13.6 GBQ2_K637details
Ornith 1.5 35B A3B36B (3.0B act.)Q2_K13.2 GBQ2_K657details
Qwen3.5 35B A3B36B (3.0B act.)Q2_K13.6 GBQ2_K637details
Axion 1.0 35B35B (3.9B act.)Q2_K13.3 GBQ2_K490details
Qwen AgentWorld 35B A3B35B (3.0B act.)Q2_K13.1 GBQ2_K637details
Yi 1.5 34B Chat34BQ2_K14.3 GBno56details
Nous Hermes 2 Yi 34B34BQ2_K14.8 GBno54details
DeepSeek R1 Distill Qwen 32B33BQ2_K14.2 GBno58details
Qwen3 32B33BQ2_K14.1 GBno58details
GLM Z1 32B 041433BQ2_K12.6 GBQ2_K58details
GLM 4 32B 041433BQ2_K12.6 GBQ2_K58details
granite 4.0 h small32B (8.8B act.)Q2_K11.7 GBQ2_K223details
gemma 4 31B it31BQ2_K13.9 GBno61details
GLM 4.7 Flash31B (4.6B act.)Q2_K11.7 GBQ2_K428details
Qwen3 Coder 30B A3B Instruct31B (3.0B act.)Q2_K11.9 GBQ2_K648details
North Mini Code 1.030B (3.3B act.)Q2_K12.0 GBQ2_K578details
Hy Memory 30B A3B30B (3.0B act.)Q2_K11.9 GBQ2_K637details
Muse Glimmer 30B30BQ2_K11.4 GBQ2_K64details
granite 4.2 30b29BQ2_K12.7 GBno66details
Qwen3.8 27B28BQ2_K11.1 GBQ2_K69details
Qwen3.6 27B28BQ2_K11.1 GBQ2_K69details
Qwen3.5 27B28BQ2_K11.1 GBQ2_K69details
gemma 3 27b it27BQ2_K11.7 GBQ2_K68details
Arete 27B27BQ3_K_M13.6 GBQ2_K54details
MiniMax H3 Image PE27BQ2_K11.3 GBQ2_K67details
Trinity Mini26B (2.0B act.)Q3_K_M12.3 GBQ3_K_M788details
GLiDE 27B26BQ3_K_M13.2 GBQ2_K56details
diffusiongemma 26B A4B it26B (4.0B act.)Q3_K_M13.2 GBQ2_K367details
gemma 4 26B A4B it26B (4.0B act.)Q3_K_M13.3 GBQ2_K363details
LFM2.5 24B A2B24B (2.0B act.)Q3_K_M11.7 GBQ3_K_M734details
ERNIE 4.5 21B A3B PT22BQ4_K_M13.5 GBQ3_K_M54details
ERNIE 4.5 21B A3B Thinking22BQ4_K_M13.5 GBQ3_K_M54details
gpt oss safeguard 20b22B (4.8B act.)Q8_012.3 GBQ8_0266details
gpt oss 20b21B (4.2B act.)Q8_012.3 GBQ8_0295details
Groot 19B A3B 26102319B (3.0B act.)Q4_K_M12.7 GBQ4_K_M372details
Kimi VL A3B Thinking 250616B (3.0B act.)Q6_K13.5 GBQ5_K_M291details
Kimi VL A3B Instruct16B (3.0B act.)Q5_K_M12.2 GBQ5_K_M326details
Kimi VL A3B Thinking16B (3.0B act.)Q6_K13.5 GBQ5_K_M291details
deepseek moe 16b base16B (2.8B act.)Q4_K_M12.5 GBQ2_K383details
LLaDA2.1 mini16B (1.4B act.)Q6_K13.4 GBQ5_K_M610details
Moonlight 16B A3B Instruct16B (3.0B act.)Q5_K_M12.2 GBQ5_K_M317details
DeepSeek Coder V2 Lite Instruct16B (2.7B act.)Q5_K_M11.9 GBQ5_K_M357details
DeepSeek V2 Lite16B (2.7B act.)Q5_K_M11.3 GBQ5_K_M378details
DeepSeek V2 Lite Chat16B (2.7B act.)Q5_K_M11.9 GBQ5_K_M357details
Qwen2.5 Coder 14B Instruct15BQ6_K13.5 GBQ3_K_M59details
DeepSeek R1 Distill Qwen 14B15BQ6_K13.5 GBQ3_K_M59details
Qwen3 14B15BQ6_K13.2 GBQ3_K_M59details
phi 415BQ6_K13.4 GBQ3_K_M60details
gemma 3 12b it12BQ8_013.4 GBQ6_K57details
Llama Guard 4 12B12BQ6_K11.5 GBQ4_K_M73details
Llama 3.2 11B Vision Instruct11BQ8_011.6 GBQ5_K_M69details
Llama 3.2 11B Vision11BQ8_012.4 GBQ5_K_M63details
GLM 4.6V Flash10BQ8_010.3 GBQ8_072details
GLM 4.1V 9B Thinking10BQ8_010.3 GBQ8_072details
Step3 VL 10B10BQ8_09.9 GBQ8_082details
SenseNova Loop 9B Preview9.7BQ8_010.0 GBQ8_075details
Qwen3.5 9B9.7BQ8_010.0 GBQ8_075details
Ornith 1.5 9B9.7BQ8_010.0 GBQ8_075details
Qwen3.5 9B Base9.7BQ8_010.0 GBQ8_075details
MiMo V2.6 Distill Qwen 9B9.4BQ8_010.0 GBQ8_075details
glm 4 9b9.4BQ6_K13.4 GBno87details
internlm3 8b instruct8.8BQ8_09.7 GBQ8_077details
granite 4.2 8b8.8BQ8_010.5 GBQ6_K77details
granite 4.1 8b8.8BQ8_010.5 GBQ6_K77details
Qwen3 VL 8B Instruct8.8BQ8_09.9 GBQ8_082details
Molmo2 8B8.7BQ8_09.9 GBQ8_082details
LFM2.5 8B A1B8.5B (1.0B act.)Q8_09.1 GBQ8_0674details
Qwen2.5 VL 7B Instruct8.3BQ8_08.7 GBQ8_089details
UI TARS 1.5 7B8.3BQ8_08.7 GBQ8_089details
olmOCR 2 7B 10258.3BQ8_08.7 GBQ8_089details
olmOCR 7B 0225 preview8.3BQ8_08.7 GBQ8_089details
UI TARS 7B DPO8.3BQ8_08.7 GBQ8_089details
Qwen3 8B8.2BQ8_09.9 GBQ8_082details
DeepSeek R1 0528 Qwen3 8B8.2BQ8_09.9 GBQ8_082details
granite 3.3 8b instruct8.2BQ8_09.8 GBQ8_083details
granite 3.1 8b instruct8.2BQ8_09.8 GBQ8_083details
granite 3.2 8b instruct8.2BQ8_09.8 GBQ8_083details
Apertus 8B Instruct 25098.1BQ8_09.6 GBQ8_084details
Llama 3.1 8B Instruct8.0BQ8_09.6 GBQ8_084details
Llama 3.1 8B8.0BQ8_09.6 GBQ8_084details
Hermes 3 Llama 3.1 8B8.0BQ8_09.6 GBQ8_084details
DeepSeek R1 Distill Llama 8B8.0BQ8_09.6 GBQ8_084details
Meta Llama 3 8B Instruct8.0BQ8_09.6 GBQ8_084details
Meta Llama Guard 2 8B8.0BQ8_09.6 GBQ8_084details
Meta Llama 3.1 8B Instruct8.0BQ8_09.6 GBQ8_084details
Ling 3.0 tiny7.9B (1.4B act.)Q8_08.7 GBQ8_0487details
MiMo 7B RL7.8BQ8_09.4 GBQ8_088details
internlm2 5 7b chat7.7BQ8_09.2 GBQ8_087details
internlm2 chat 7b7.7BQ8_09.2 GBQ8_087details
Qwen2.5 7B Instruct7.6BQ8_08.7 GBQ8_089details
Qwen2.5 Coder 7B Instruct7.6BQ8_08.7 GBQ8_089details
DeepSeek R1 Distill Qwen 7B7.6BQ8_08.7 GBQ8_089details
Falcon H1R 7B7.6BQ8_08.5 GBQ8_089details
Olmo 3 7B Instruct7.3BQ8_010.3 GBQ8_092details
Olmo 3 1025 7B7.3BQ8_010.3 GBQ8_092details
Olmo 3 7B Think7.3BQ8_010.3 GBQ8_092details
Mistral 7B v0.17.2BQ8_08.6 GBQ8_093details
Mistral 7B Instruct v0.27.2BQ8_08.6 GBQ8_093details
Mistral 7B Instruct v0.17.2BQ8_08.6 GBQ8_093details
falcon 7b7.2BQ8_012.1 GBno94details
falcon 7b instruct7.2BQ8_011.8 GBno97details
granite 4.0 h tiny6.9B (1.5B act.)F1613.6 GBQ8_0244details
OLMoE 1B 7B 09246.9B (1.3B act.)Q8_08.4 GBQ8_0526details
deepseek coder 7b instruct v1.56.9BQ8_011.2 GBno98details
deepseek coder 6.7b instruct6.7BQ8_011.1 GBno100details
deepseek coder 6.7b base6.7BQ8_011.1 GBno100details
Llama 2 7b hf6.7BQ8_011.1 GBno100details
Yi 6B 200K6.1BF1612.3 GBQ8_059details
Yi 6B6.1BF1612.3 GBQ8_059details
Molmo2 4B4.9BF1610.0 GBF1681details
WeVisOne VL 4B4.8BF1610.0 GBF1681details
Mage VL Cowork4.7BF1610.0 GBF1681details
Qianfan OCR4.7BF1610.1 GBF1681details
Qwen3.5 4B4.7BF169.0 GBF1685details
S5 nano4.7BF169.2 GBF1683details
LightOnOCR 3 4B4.5BF169.0 GBF1685details
SingVL S1 4B4.5BF169.0 GBF1685details
Intern Decision 4B4.5BF169.0 GBF1685details
Qwen3 VL 4B Instruct4.4BF169.3 GBF1689details
gemma 3 4b it4.3BF168.4 GBF1692details
medgemma 1.5 4b it4.3BF168.4 GBF1692details
medgemma 4b it4.3BF168.4 GBF1692details
gemma 3 4b pt4.3BF169.2 GBF1683details
Phi 3.5 vision instruct4.1BF1611.2 GBQ4_K_M86details
Qwen3 4B4.0BF169.3 GBF1689details
Qwen3 4B Instruct 25074.0BF169.3 GBF1689details
granite vision 4.1 4b4.0BF167.6 GBF16105details
Phi 4 mini instruct3.8BF168.9 GBF1693details
Phi 3 mini 4k instruct3.8BF1610.6 GBQ5_K_M94details
Phi 3.5 mini instruct3.8BF1610.6 GBQ5_K_M94details
Phi 3 mini 128k instruct3.8BF1610.6 GBQ5_K_M94details
Qwen2.5 VL 3B Instruct3.8BF166.7 GBF16116details
granite 4.2 3b3.7BF168.0 GBF1698details
granite 4.1 3b llmon preview3.7BF168.0 GBF1698details
Raon 1 nano3.5BF167.0 GBF16114details
granite 4.1 3b3.4BF167.6 GBF16105details
granite 4.0 micro3.4BF167.6 GBF16105details
DeepSeek OCR 23.4B (1.2B act.)F166.6 GBF16346details
deepseek vl2 tiny3.4B (1.2B act.)F167.4 GBF16305details
Unlimited OCR 23.3B (1.1B act.)F166.6 GBF16355details
Unlimited OCR3.3B (1.1B act.)F166.6 GBF16357details
DeepSeek OCR3.3B (1.1B act.)F166.6 GBF16357details
granite 3.1 3b a800m instruct3.3B (883M act.)F167.1 GBF16406details
Llama 3.2 3B Instruct3.2BF1613.5 GBQ8_056details
Llama 3.2 3B3.2BF167.5 GBF16112details
Hermes 3 Llama 3.2 3B3.2BF167.5 GBF16111details
d1 3B3.1BF1610.8 GBF1666details
LFM2.5 VL 3B3.1BF1610.8 GBF1666details
Qwen2.5 3B Instruct3.1BF166.7 GBF16116details
SmolLM3 3B3.1BF166.9 GBF16116details
SmolLM3 3B Base3.1BF166.9 GBF16117details
phi 22.8BF168.2 GBQ8_0129details
LFM2.5 2.6B2.7BF1610.8 GBF1666details
granite 3.3 2b instruct2.5BF165.8 GBF16141details
MiniCPM5 2B2.5BF165.7 GBF16142details
WeVisOne VL 2B2.4BF165.4 GBF16176details
Qwen3.5 2B2.3BF164.5 GBF16190details
SmolVLM Instruct2.2BF165.4 GBF16198details
Qwen3 1.7B2.0BF164.8 GBF16208details
Qwen3.8 27B DSpine b162.0BF164.8 GBF16177details
d1 2B1.9BF167.2 GBF16103details
LFM3 VL 2B1.9BF167.2 GBF16103details
DeepSeek R1 Distill Qwen 1.5B1.8BF164.2 GBF16201details
LFM3 1.7B1.7BF167.2 GBF16103details
LFM3 1.7B Base1.7BF164.0 GBF16207details
SmolLM2 1.7B Instruct1.7BF165.2 GBF16209details
SmolLM2 1.7B1.7BF165.2 GBF16209details
SmolLM 1.7B1.7BF165.2 GBF16209details
granite 4.0 1b1.6BF164.3 GBF16219details
Qwen2.5 1.5B Instruct1.5BF163.8 GBF16232details
OLMo 2 0425 1B1.5BF164.4 GBF16241details
deepseek coder 1.3b instruct1.3BF164.5 GBF16266details
Llama 3.2 1B1.2BF163.2 GBF16290details
Llama 3.2 1B Instruct1.2BF165.5 GBF16145details
LFM2.5 1.2B Instruct1.2BF165.0 GBF16153details
HunyuanOCR1.1BF163.5 GBF16320details
PaddleOCR VL 1.6959MF162.5 GBF16374details
Qwen3.5 0.8B873MF162.4 GBF16473details
LightOnOCR 3 0.8B853MF162.4 GBF16473details
FastVLM 0.5B759MF162.2 GBF16472details
Qwen3 0.6B752MF162.7 GBF16598details
GOT OCR2 0716MF162.8 GBF16501details
Nemotron 3.5 Parse631MF161.9 GBF16549details
GOT OCR 2.0 hf561MF162.5 GBF16640details
SmolVLM2 500M Video Instruct507MF161.6 GBF16874details
SmolVLM 500M Instruct507MF161.6 GBF16874details
Qwen2.5 0.5B Instruct494MF161.7 GBF16726details
LFM2.5 VL 450M449MF161.9 GBF16504details
SmolLM2 360M Instruct362MF161.5 GBF16988details
SmolLM2 360M362MF161.5 GBF16988details
LFM2.5 350M354MF161.9 GBF16504details
LFM2 350M354MF161.4 GBF161011details

Just too big at 4-bit

These need up to 35% more memory than the RTX 4080 has at Q4_K_M and 8K context. A smaller quant or shorter context may work.

ModelQ4_K_M needsShort by
Trinity Mini15.8 GB0.6 GB
GLiDE 27B16.2 GB1.0 GB
MiniMax H3 Image PE16.8 GB1.6 GB
Qwen3.8 27B16.8 GB1.6 GB
Qwen3.5 27B17.0 GB1.8 GB
Qwen3.6 27B17.0 GB1.8 GB
diffusiongemma 26B A4B it17.1 GB1.9 GB
gemma 4 26B A4B it17.2 GB2.0 GB
gemma 3 27b it17.3 GB2.1 GB
Arete 27B17.4 GB2.2 GB
Muse Glimmer 30B17.9 GB2.7 GB
GLM 4.7 Flash18.2 GB3.0 GB
Hy Memory 30B A3B18.5 GB3.3 GB
Qwen3 Coder 30B A3B Instruct18.7 GB3.5 GB
granite 4.0 h small18.9 GB3.7 GB

Questions

What is the biggest model that runs on the RTX 4080?

Of the models we track, the largest that fits at 8K context is flint (40B) at Q2_K, needing about 14.8 GB.

How much of the RTX 4080's memory can a model use?

About 15.2 GB: we assume 95% of the 16 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.