Skip to content

What can the L40S run?

What fits on the L40S
Text models: 184Visual models: 12Strongest first
Memory
48 GB
Bandwidth
864 GB/s
Maker
NVIDIA

Open the calculator on this machineL40S in our inventoryEvery model for 48 GB of VRAM

Text models that run on it

Text models that run on the L40S, best first
ModelBest fitContextSpeedStrongest at
Qwen3.8 27B8-bit256K21 tok/sChat, Coding, Agents, Docs
Muse Glimmer 30B8-bit128K19 tok/sDocs
Qwen3.5 122B A10B2-bit32K107 tok/sChat, Coding, Agents, Docs
Nemotron 3 Super 120B A12B2-bit184K77 tok/sChat, Coding, Agents, Docs
Nemotron 3.5 Bolt 74B A9B4-bit256K69 tok/sChat, Coding, Agents, Docs
Qwen3.6 35B A3B8-bit256K129 tok/sChat, Coding, Agents, Docs
Gemma 4 31B8-bit256K18 tok/sChat, Coding, Agents, Docs
Gemma 4 26B A4B8-bit256K118 tok/sChat, Coding, Agents, Docs
GLM-4.7 Flash8-bit198K124 tok/sChat, Coding, Agents
Gemma 4 12B8-bit256K45 tok/sChat, Coding, Docs
Mistral Small 4 119B2-bit136K162 tok/sChat, Docs
gpt-oss 120B2-bit104K188 tok/sChat, Coding
Qwen3 Coder Next 80B3-bit256K184 tok/sCoding, Agents
Nemotron 3.5 Lightning 30B A3B8-bit256K118 tok/sAgents, Docs
Devstral Small 2 24BFP8128K23 tok/sCoding
LFM2.5 24B A2B8-bit128K169 tok/sAgents
Mistral Small 3.2 24B8-bit128K24 tok/sChat, Coding, Agents
Ministral 3 14B8-bit192K41 tok/sCoding, Agents, Docs
gpt-oss 20BMXFP4128K176 tok/sAgents
Ministral 3 8B8-bit256K64 tok/sDocs
Llama 3.3 70B3-bit36K17 tok/sChat, Coding, Agents, Docs
Olmo 3.1 32B8-bit64K17 tok/sChat, Coding, Agents, Docs
Llama 3.1 8B8-bit128K71 tok/sChat, Coding, Agents, Docs
Qwen3 14B8-bit40K39 tok/sNot rated
Qwen3 VL 8B Instruct8-bit256K65 tok/sText and images
Qwen3 8B8-bit40K70 tok/sNot rated
Qwen3 32B8-bit40K17 tok/sNot rated
Qwen3.6 27B8-bit256K21 tok/sText and images
Qwen2.5 Coder 14B Instruct8-bit32K38 tok/sCoding
Qwen3 Coder 30B A3B Instruct8-bit152K123 tok/sCoding
Qwen3.5 27B8-bit256K21 tok/sText and images
Qwen3.5 35B A3B8-bit256K129 tok/sText and images
NVIDIA Nemotron 3 Nano 30B A3B BF168-bit256K118 tok/sNot rated
GLM 4.5 Air2-bit32K83 tok/sNot rated
Qwen3 30B A3B8-bit40K123 tok/sNot rated
Qwen3 30B A3B Instruct 25078-bit152K123 tok/sNot rated
Laguna XS.28-bit256K130 tok/sNot rated
OTel 2.0 LLM 31B IT8-bit256K18 tok/sNot rated
Qwen2.5 32B Instruct8-bit32K17 tok/sNot rated
MiniCPM V 4.7 35B A3B8-bit256K129 tok/sText and images
Qwen2.5 14B Instruct8-bit32K38 tok/sNot rated
Laguna XS 2.18-bit256K130 tok/sNot rated
Diffusiongemma 26B A4B IT8-bit256K113 tok/sText and images
Qwen2.5 VL 32B Instruct8-bit44K17 tok/sText and images
Qwen2.5 Coder 32B Instruct8-bit32K17 tok/sCoding
Laguna S 2.12-bit72K123 tok/sNot rated
Qwen3 VL 30B A3B Instruct8-bit144K123 tok/sText and images
EXAONE 3.5 7.8B Instruct8-bit32K72 tok/sNot rated
DeepSeek R1 0528 Qwen3 8B8-bit128K70 tok/sNot rated
Gemma 3 12B IT8-bit128K47 tok/sText and images
Phi 3.5 Mini Instruct8-bit104K133 tok/sNot rated
DeepSeek R1 Distill Qwen 32B8-bit48K17 tok/sNot rated
Gemma 3 27B IT8-bit128K21 tok/sText and images
Ornith 1.5 35B A3B8-bit256K129 tok/sNot rated
Ornith 1.5 122B2-bit32K91 tok/sNot rated
GPT Neox 20B8-bit2K27 tok/sNot rated
Apertus 8B Instruct 25098-bit64K71 tok/sNot rated
Phi 48-bit16K38 tok/sNot rated
Phi 3 Mini 128K Instruct8-bit104K133 tok/sNot rated
Qwen3 VL 32B Instruct8-bit44K17 tok/sText and images

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 184 are shown.

Image and video models

Image and video models that run on the L40S
ModelBest fitPer outputStrongest at
MiniMax H316-bit3.0 minVideo
LTX-2.516-bit36 sVideo
FLUX.1 [dev]16-bit5.9 sImages
Qwen-Image16-bit1.6 minImages
FLUX.2 [dev]8-bit43 sImages
Wan2.2 I2V A14B16-bit14 minVideo
Wan2.1 I2V 14B16-bit14 minVideo
Wan 2.1 T2V-14B16-bit18 minVideo
Krea 2 Turbo16-bit5.9 sImages
Qwen Image 251216-bit1.6 minImages
Wan 2.2 A14B16-bit14 minVideo
HunyuanVideo 1.516-bit17 minVideo

Only just too big

These need up to a quarter more memory than the L40S has. Spilling into system RAM or a lower precision can get them running, slowly.

Questions

What is the best LLM for the L40S?

Of the open models we track, the strongest that run well on the L40S are Qwen3.8 27B, Muse Glimmer 30B, Qwen3.5 122B A10B. 184 text models fit it in all.

Calculated from published specs, not benchmarked. All machines