RunsOnMy ModelsGPUs & MacsBy VRAMBest GPUCan I run…?OllamaStart hereAPI pricesVRAM calculatorNew

What AI models run on the L40S?

The L40S has 48 GB of VRAM (864 GB/s). About 45.6 GB is usable for a model. 225 of 298 open models we track fit at 8K context.

225models fit (8K ctx)
222still fit at 32K ctx
127Blargest that fits
864 GB/smemory bandwidth

This card has 48 GB of VRAM. See every model that fits in 48 GB VRAM. Other 48 GB hardware: RTX A6000, RTX 6000 Ada, Mac M4 Pro 48GB.

Using a specific tool? Ollama models that run on the L40S

Models that fit, largest first

ModelParamsBest quant (8K)Needs32K contextMax tok/s (est.)Details
Ling 3.0 Flash127B (5.6B act.)Q2_K43.8 GBQ2_K428details
Ling 3.0 Flash VL125B (5.9B act.)Q2_K45.6 GBno380details
GLM 4.5 Air110B (15B act.)Q2_K45.1 GBno138details
Llama 4 Scout 17B 16E Instruct109B (18B act.)Q2_K40.2 GBQ2_K128details
GLM 4.5V108B (14B act.)Q2_K39.7 GBQ2_K161details
GLM 4.6V108B (14B act.)Q2_K43.8 GBno146details
Ling 2.6 Flash107B (7.6B act.)Q2_K38.5 GBQ2_K302details
Hunyuan A13B Instruct80B (13B act.)Q3_K_M36.4 GBQ3_K_M143details
Kimi Dev 72B73BQ3_K_M38.3 GBQ2_K23details
Apertus v1.5 70B72BQ3_K_M36.2 GBQ2_K25details
Llama 3.3 70B Instruct71BQ3_K_M35.1 GBQ2_K25details
Llama 3.1 70B Instruct71BQ3_K_M35.1 GBQ2_K25details
Meta Llama 3 70B Instruct71BQ3_K_M35.1 GBQ2_K25details
DeepSeek R1 Distill Llama 70B71BQ3_K_M35.1 GBQ2_K25details
Llama 3.1 70B71BQ3_K_M35.1 GBQ2_K25details
Hermes 4 70B71BQ3_K_M35.1 GBQ2_K25details
Kimi Linear 48B A3B Instruct49B (3.0B act.)Q6_K38.6 GBQ6_K350details
Nous Hermes 2 Mixtral 8x7B DPO47B (13B act.)Q6_K37.2 GBQ6_K82details
Falcon 40B42BQ6_K33.6 GBQ6_K25details
Flint40B (1.3B act.)Q8_040.4 GBQ8_0649details
Maxwell V1 Flash37B (2.3B act.)Q8_036.9 GBQ8_0365details
Hermes 4.3 36B36BQ8_038.5 GBQ6_K22details
Seed OSS 36B Instruct36BQ8_038.5 GBQ6_K22details
Qwen3.6 35B A3B36B (3.0B act.)Q8_036.2 GBQ8_0274details
Ornith 1.5 35B A3B36B (3.0B act.)Q8_036.2 GBQ8_0274details
Qwen3.5 35B A3B36B (3.0B act.)Q8_036.2 GBQ8_0274details
Sage A1.536B (4.7B act.)Q8_036.6 GBQ8_0171details
Qwen AgentWorld 35B A3B35B (3.0B act.)Q8_035.4 GBQ8_0270details
Yi 34B 200K34BQ8_036.4 GBQ6_K24details
Yi 34B34BQ8_036.4 GBQ6_K24details
Yi 34B Chat34BQ8_036.4 GBQ6_K24details
Yi 1.5 34B Chat34BQ8_036.4 GBQ6_K24details
Nous Hermes 2 Yi 34B34BQ8_036.4 GBQ6_K24details
DeepSeek R1 Distill Qwen 32B33BQ8_035.1 GBQ6_K25details
Qwen3 32B33BQ8_035.1 GBQ6_K25details
GLM Z1 32B 041433BQ8_033.4 GBQ8_025details
GLM 4 32B 041433BQ8_033.4 GBQ8_025details
Olmo 3 32B Think32BQ8_033.7 GBQ8_025details
Granite 4.0 H Small32B (8.8B act.)Q8_032.6 GBQ8_092details
Gemma 4 31B IT31BQ8_033.3 GBQ8_026details
Gemma 4 31B31BQ8_033.3 GBQ8_026details
GLM 4.7 Flash31B (4.6B act.)Q8_030.8 GBQ8_0184details
Qwen3 Coder 30B A3B Instruct31B (3.0B act.)Q8_031.7 GBQ8_0271details
North Mini Code 1.030B (3.3B act.)Q8_031.6 GBQ8_0246details
Hy Memory 30B A3B30B (3.0B act.)Q8_031.1 GBQ8_0271details
Muse Glimmer 30B30BQ8_030.4 GBQ8_027details
Granite 4.2 30B29BQ8_031.6 GBQ8_028details
Raon 1 Mini Preview29B (4.3B act.)Q8_030.4 GBQ8_0183details
Qwen3.8 27B28BQ8_028.5 GBQ8_030details
Qwen3.6 27B28BQ8_028.5 GBQ8_030details
Qwen3.5 27B28BQ8_028.5 GBQ8_030details
Gemma 3 27B IT27BQ8_028.7 GBQ8_030details
Arete 27B27BQ8_028.1 GBQ8_030details
MiniMax H3 Image PE27BQ8_028.0 GBQ8_030details
Trinity Mini26B (2.0B act.)Q8_026.9 GBQ8_0413details
GLiDE 27B26BQ8_027.2 GBQ8_031details
DiffusionGemma 26B A4B IT26B (4.0B act.)Q8_026.4 GBQ8_0208details
Gemma 4 26B A4B IT26B (4.0B act.)Q8_026.4 GBQ8_0208details
LFM2.5 24B A2B24B (2.0B act.)Q8_024.5 GBQ8_0406details
ERNIE 4.5 21B A3B PT22BQ8_022.6 GBQ8_037details
ERNIE 4.5 21B A3B Thinking22BQ8_022.6 GBQ8_037details
gpt-oss Safeguard 20B22B (4.8B act.)F1613.8 GBF16281details
gpt-oss 20B21B (4.2B act.)F1613.8 GBF16312details
Kimi VL A3B Thinking 250616B (3.0B act.)F1630.7 GBF16148details
Kimi VL A3B Thinking16B (3.0B act.)F1631.5 GBF16144details
Kimi VL A3B Instruct16B (3.0B act.)F1631.5 GBF16144details
DeepSeek MoE 16B Base16B (2.8B act.)F1632.8 GBF16153details
LLaDA2.1 Mini16B (1.4B act.)F1631.3 GBF16301details
Moonlight 16B A3B Instruct16B (3.0B act.)F1630.7 GBF16144details
DeepSeek Coder V2 Lite Instruct16B (2.7B act.)F1630.1 GBF16162details
DeepSeek V2 Lite16B (2.7B act.)F1630.1 GBF16162details
DeepSeek V2 Lite Chat16B (2.7B act.)F1630.1 GBF16162details
Qwen2.5 Coder 14B Instruct15BF1629.7 GBF1629details
DeepSeek R1 Distill Qwen 14B15BF1629.7 GBF1629details
Qwen3 14B15BF1629.4 GBF1629details
Phi 415BF1629.5 GBF1629details
Gemma 3 12B IT12BF1623.6 GBF1637details
Llama Guard 4 12B12BF1624.6 GBF1636details
KORMo Web 11B11BF1622.7 GBF1639details
Llama 3.2 11B Vision Instruct11BF1620.1 GBF1644details
Llama 3.2 11B Vision11BF1621.7 GBF1641details
GLM 4.6V Flash10BF1618.5 GBF1646details
GLM 4.1V 9B Thinking10BF1618.5 GBF1646details
Step3 VL 10B10BF1617.1 GBF1653details
SenseNova Loop 9B Preview9.7BF1617.8 GBF1648details
Qwen3.5 9B9.7BF1618.3 GBF1647details
Ornith 1.5 9B9.7BF1617.8 GBF1648details
SenseNova Loop 9B9.4BF1618.7 GBF1646details
MiMo V2.6 Distill Qwen 9B9.4BF1617.8 GBF1648details
GLM 4 9B9.4BF1618.5 GBF1646details
Apertus v1.5 8B8.9BF1618.2 GBF1649details
InternLM3 8B Instruct8.8BF1617.4 GBF1649details
Granite 4.2 8B8.8BF1618.2 GBF1649details
Granite 4.1 8B8.8BF1618.2 GBF1649details
Granite 4.1 8B Llmon Preview8.8BF1618.2 GBF1649details
Qwen3 VL 8B Instruct8.8BF1617.1 GBF1653details
Molmo2 8B8.7BF1617.1 GBF1653details
LFM2.5 8B A1B8.5B (1.0B act.)F1616.5 GBF16432details
Qwen2.5 VL 7B Instruct8.3BF1615.3 GBF1657details
UI TARS 1.5 7B8.3BF1615.3 GBF1657details
olmOCR 2 7B 10258.3BF1615.3 GBF1657details
olmOCR 7B 0225 Preview8.3BF1616.6 GBF1652details
UI TARS 7B DPO8.3BF1615.3 GBF1657details
Qwen3 8B8.2BF1617.1 GBF1653details
DeepSeek R1 0528 Qwen3 8B8.2BF1617.1 GBF1653details
Granite 3.3 8B Instruct8.2BF1617.0 GBF1653details
Granite 3.1 8B Instruct8.2BF1617.0 GBF1653details
Granite 3.2 8B Instruct8.2BF1617.0 GBF1653details
Apertus 8B Instruct 25098.1BF1616.7 GBF1654details
Llama 3.1 8B Instruct8.0BF1616.6 GBF1654details
Llama 3.1 8B8.0BF1616.6 GBF1654details
Hermes 3 Llama 3.1 8B8.0BF1616.6 GBF1654details
Llama Guard 3 8B8.0BF1616.6 GBF1654details
DeepSeek R1 Distill Llama 8B8.0BF1616.6 GBF1654details
Meta Llama 3 8B Instruct8.0BF1616.6 GBF1654details
Meta Llama Guard 2 8B8.0BF1616.6 GBF1654details
Meta Llama 3.1 8B Instruct8.0BF1616.6 GBF1654details
Aya Expanse 8B8.0BF1616.9 GBF1654details
Ling 3.0 Tiny7.9B (1.4B act.)F1615.6 GBF16312details
MiMo 7B RL7.8BF1616.4 GBF1655details
InternLM2.5 7B Chat7.7BF1616.0 GBF1656details
InternLM2 Chat 7B7.7BF1616.0 GBF1656details
Qwen2.5 7B Instruct7.6BF1615.3 GBF1657details
DeepSeek R1 Distill Qwen 7B7.6BF1615.3 GBF1657details
Qwen2.5 Coder 7B Instruct7.6BF1615.3 GBF1657details
Olmo 3 7B Instruct7.3BF1616.7 GBF1659details
Olmo 3 1025 7B7.3BF1616.7 GBF1659details
Olmo 3 7B Think7.3BF1616.7 GBF1659details
Mistral 7B v0.17.2BF1614.9 GBF1660details
Mistral 7B Instruct v0.27.2BF1614.9 GBF1660details
Mistral 7B Instruct v0.17.2BF1614.9 GBF1660details
Falcon 7B7.2BF1614.0 GBF1660details
Falcon 7B Instruct7.2BF1614.0 GBF1660details
Granite 4.0 H Tiny6.9B (1.5B act.)F1613.6 GBF16295details
OLMoE 1B 7B 09246.9B (1.3B act.)F1614.4 GBF16337details
DeepSeek Coder 7B Instruct v1.56.9BF1617.2 GBF1663details
DeepSeek Coder 6.7B Instruct6.7BF1617.0 GBF1664details
DeepSeek Coder 6.7B Base6.7BF1617.0 GBF1664details
Llama 2 7B HF6.7BF1617.0 GBF1664details
Yi 6B 200K6.1BF1612.3 GBF1671details
Yi 6B6.1BF1612.3 GBF1671details
Molmo2 4B4.9BF1610.0 GBF1698details
WeVisOne VL 4B4.8BF1610.0 GBF1698details
Mage VL Cowork4.7BF1610.0 GBF1698details
Qianfan OCR4.7BF1610.1 GBF1698details
Qwen3.5 4B4.7BF169.2 GBF16100details
SingVL S1 4B4.5BF169.0 GBF16103details
LightOnOCR 3 4B4.5BF169.0 GBF16103details
Intern Decision 4B4.5BF169.0 GBF16103details
Qwen3 VL 4B Instruct4.4BF169.3 GBF16107details
Gemma 3 4B IT4.3BF168.4 GBF16111details
MedGemma 1.5 4B IT4.3BF168.4 GBF16111details
MedGemma 4B IT4.3BF168.4 GBF16111details
TranslateGemma 4B IT4.3BF169.2 GBF16100details
Gemma 3 4B PT4.3BF168.4 GBF16111details
Phi 3.5 Vision Instruct4.1BF1611.2 GBF16104details
Qwen3 4B4.0BF169.3 GBF16107details
Qwen3 4B Instruct 25074.0BF169.3 GBF16107details
Granite Vision 4.1 4B4.0BF167.6 GBF16127details
Phi 4 Mini Instruct3.8BF168.9 GBF16113details
Phi 3 Mini 4K Instruct3.8BF1610.6 GBF16113details
Phi 3.5 Mini Instruct3.8BF1610.6 GBF16113details
Phi 3 Mini 128K Instruct3.8BF1610.6 GBF16113details
Qwen2.5 VL 3B Instruct3.8BF168.0 GBF16115details
Granite 4.2 3B3.7BF168.0 GBF16118details
Granite 4.1 3B Llmon Preview3.7BF168.0 GBF16118details
Raon 1 Nano3.5BF167.0 GBF16138details
Granite 4.1 3B3.4BF167.6 GBF16127details
Granite 4.0 Micro3.4BF167.6 GBF16127details
DeepSeek OCR 23.4B (1.2B act.)F166.6 GBF16417details
DeepSeek VL2 Tiny3.4B (1.2B act.)F167.4 GBF16368details
Tiny Aya Global3.3BF167.5 GBF16129details
Unlimited OCR 23.3B (1.1B act.)F166.6 GBF16427details
Unlimited OCR3.3B (1.1B act.)F166.6 GBF16430details
DeepSeek OCR3.3B (1.1B act.)F166.6 GBF16430details
Granite 3.1 3B A800M Instruct3.3B (883M act.)F167.1 GBF16489details
Llama 3.2 3B Instruct3.2BF167.5 GBF16134details
Llama 3.2 3B3.2BF167.5 GBF16134details
Hermes 3 Llama 3.2 3B3.2BF167.5 GBF16134details
d1 3B3.1BF165.8 GBF16160details
LFM2.5 VL 3B3.1BF165.8 GBF16160details
Qwen2.5 3B Instruct3.1BF166.7 GBF16140details
SmolLM3 3B3.1BF166.9 GBF16140details
SmolLM3 3B Base3.1BF166.9 GBF16140details
Phi 22.8BF168.2 GBF16155details
LFM2.5 2.6B2.7BF1610.8 GBF1680details
Granite 3.3 2B Instruct2.5BF165.8 GBF16170details
MiniCPM5 2B2.5BF165.7 GBF16171details
WeVisOne VL 2B2.4BF165.4 GBF16212details
Cosmos Reason2 2B2.4BF165.4 GBF16212details
Qwen3.5 2B2.3BF164.6 GBF16222details
SmolVLM Instruct2.2BF165.4 GBF16238details
Qwen3 1.7B2.0BF165.4 GBF16212details
d1 2B1.9BF164.0 GBF16249details
LFM3 VL 2B1.9BF164.0 GBF16249details
DeepSeek R1 Distill Qwen 1.5B1.8BF164.2 GBF16243details
LFM3 1.7B1.7BF167.2 GBF16125details
LFM3 1.7B Base1.7BF164.0 GBF16249details
SmolLM2 1.7B Instruct1.7BF165.2 GBF16252details
SmolLM2 1.7B1.7BF165.2 GBF16252details
SmolLM 1.7B1.7BF165.2 GBF16252details
Granite 4.0 1B1.6BF164.3 GBF16264details
Qwen2.5 1.5B Instruct1.5BF163.8 GBF16280details
OLMo 2 0425 1B1.5BF164.4 GBF16291details
DeepSeek Coder 1.3B Instruct1.3BF164.5 GBF16321details
Llama 3.2 1B1.2BF163.2 GBF16348details
Llama 3.2 1B Instruct1.2BF163.2 GBF16348details
LFM2.5 1.2B Instruct1.2BF165.0 GBF16184details
LFM2 1.2B1.2BF163.0 GBF16369details
HunyuanOCR1.1BF163.5 GBF16386details
PaddleOCR VL 1.6959MF162.5 GBF16451details
Qwen3.5 0.8B873MF162.4 GBF16555details
LightOnOCR 3 0.8B853MF162.4 GBF16570details
FastVLM 0.5B759MF162.2 GBF16569details
Qwen3 0.6B752MF163.0 GBF16572details
GOT OCR2.0716MF162.8 GBF16603details
Nemotron 3.5 Parse631MF161.9 GBF16661details
GOT OCR 2.0 HF561MF162.5 GBF16771details
SmolVLM2 500M Video Instruct507MF161.6 GBF161053details
SmolVLM 500M Instruct507MF161.6 GBF161053details
Qwen2.5 0.5B Instruct494MF161.7 GBF16874details
LFM2.5 VL 450M449MF161.3 GBF161214details
SmolLM2 360M Instruct362MF161.5 GBF161191details
SmolLM2 360M362MF161.5 GBF161191details
LFM2.5 350M354MF161.9 GBF16607details

Just too big at 4-bit

These need up to 35% more memory than the L40S has at Q4_K_M and 8K context. A smaller quant or shorter context may work.

ModelQ4_K_M needsShort by
Hunyuan A13B Instruct47.6 GB2.0 GB
gpt-oss Safeguard 120B59.5 GB13.9 GB
gpt-oss 120B59.6 GB14.0 GB
GLM 4.5V61.4 GB15.8 GB

Questions

What is the biggest model that runs on the L40S?

Of the models we track, the largest that fits at 8K context is Ling 3.0 Flash (127B) at Q2_K, needing about 43.8 GB.

How much of the L40S's memory can a model use?

About 45.6 GB: we assume 95% of the 48 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.