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

What AI models run on the RTX 3080 10GB?

The RTX 3080 10GB has 10 GB of VRAM (760 GB/s). About 9.5 GB is usable for a model. 165 of 298 open models we track fit at 8K context.

165models fit (8K ctx)
146still fit at 32K ctx
24Blargest that fits
760 GB/smemory bandwidth

This card has 10 GB of VRAM. See every model that fits in 12 GB VRAM.

Compare for local LLMs: RTX 3080 10GB vs RTX 4070 all comparisons

Using a specific tool? Ollama models that run on the RTX 3080 10GB · LM Studio · llama.cpp offload

Models that fit, largest first

ModelParamsBest quant (8K)Needs32K contextMax tok/s (est.)Details
LFM2.5 24B A2B24B (2.0B act.)Q2_K9.2 GBno1013details
ERNIE 4.5 21B A3B PT22BQ2_K8.6 GBno94details
ERNIE 4.5 21B A3B Thinking22BQ2_K8.6 GBno94details
Kimi VL A3B Thinking 250616B (3.0B act.)Q3_K_M8.4 GBQ2_K518details
Kimi VL A3B Thinking16B (3.0B act.)Q3_K_M8.4 GBQ2_K518details
Kimi VL A3B Instruct16B (3.0B act.)Q2_K7.1 GBQ2_K631details
DeepSeek MoE 16B Base16B (2.8B act.)Q2_K8.6 GBno656details
LLaDA2.1 Mini16B (1.4B act.)Q3_K_M8.4 GBQ2_K1085details
Moonlight 16B A3B Instruct16B (3.0B act.)Q2_K7.1 GBQ2_K615details
DeepSeek Coder V2 Lite Instruct16B (2.7B act.)Q3_K_M8.0 GBQ2_K584details
DeepSeek V2 Lite16B (2.7B act.)Q3_K_M8.0 GBQ2_K584details
DeepSeek V2 Lite Chat16B (2.7B act.)Q3_K_M8.4 GBQ2_K552details
Qwen2.5 Coder 14B Instruct15BQ2_K7.6 GBno132details
DeepSeek R1 Distill Qwen 14B15BQ2_K7.6 GBno132details
Qwen3 14B15BQ2_K7.1 GBno137details
Phi 415BQ2_K7.3 GBno137details
Gemma 3 12B IT12BQ4_K_M8.5 GBQ2_K104details
Llama Guard 4 12B12BQ3_K_M7.7 GBno130details
KORMo Web 11B11BQ4_K_M8.3 GBno111details
Llama 3.2 11B Vision Instruct11BQ4_K_M7.4 GBQ2_K127details
Llama 3.2 11B Vision11BQ4_K_M8.0 GBQ2_K117details
GLM 4.6V Flash10BQ5_K_M7.6 GBQ5_K_M108details
GLM 4.1V 9B Thinking10BQ5_K_M7.6 GBQ5_K_M108details
Step3 VL 10B10BQ6_K8.1 GBQ3_K_M113details
SenseNova Loop 9B Preview9.7BQ6_K8.0 GBQ5_K_M103details
Qwen3.5 9B9.7BQ6_K8.5 GBQ5_K_M95details
Ornith 1.5 9B9.7BQ6_K8.3 GBQ5_K_M99details
SenseNova Loop 9B9.4BQ6_K8.3 GBQ5_K_M98details
MiMo V2.6 Distill Qwen 9B9.4BQ6_K8.4 GBQ5_K_M98details
GLM 4 9B9.4BQ5_K_M7.7 GBQ4_K_M106details
Apertus v1.5 8B8.9BQ6_K8.4 GBQ2_K104details
InternLM3 8B Instruct8.8BQ6_K7.8 GBQ5_K_M105details
Granite 4.2 8B8.8BQ5_K_M7.7 GBQ2_K122details
Granite 4.1 8B8.8BQ5_K_M7.7 GBQ2_K122details
Granite 4.1 8B Llmon Preview8.8BQ5_K_M7.7 GBQ2_K122details
Qwen3 VL 8B Instruct8.8BQ6_K8.5 GBQ2_K106details
Molmo2 8B8.7BQ6_K8.4 GBQ2_K107details
LFM2.5 8B A1B8.5B (1.0B act.)Q6_K7.2 GBQ6_K925details
Qwen2.5 VL 7B Instruct8.3BQ6_K7.0 GBQ6_K122details
UI TARS 1.5 7B8.3BQ6_K7.0 GBQ6_K122details
olmOCR 2 7B 10258.3BQ6_K7.0 GBQ6_K122details
olmOCR 7B 0225 Preview8.3BQ6_K7.0 GBQ6_K122details
UI TARS 7B DPO8.3BQ6_K7.0 GBQ6_K122details
Qwen3 8B8.2BQ6_K8.1 GBQ2_K113details
DeepSeek R1 0528 Qwen3 8B8.2BQ6_K8.1 GBQ2_K113details
Granite 3.3 8B Instruct8.2BQ6_K8.0 GBQ2_K113details
Granite 3.1 8B Instruct8.2BQ6_K8.0 GBQ2_K113details
Granite 3.2 8B Instruct8.2BQ6_K8.0 GBQ2_K113details
Apertus 8B Instruct 25098.1BQ6_K7.8 GBQ3_K_M115details
Llama 3.1 8B Instruct8.0BQ6_K7.8 GBQ3_K_M115details
Llama 3.1 8B8.0BQ6_K7.8 GBQ3_K_M115details
Hermes 3 Llama 3.1 8B8.0BQ6_K7.8 GBQ3_K_M115details
Llama Guard 3 8B8.0BQ6_K7.8 GBQ3_K_M115details
DeepSeek R1 Distill Llama 8B8.0BQ6_K7.8 GBQ3_K_M115details
Meta Llama 3 8B Instruct8.0BQ6_K7.8 GBQ3_K_M115details
Meta Llama Guard 2 8B8.0BQ6_K7.8 GBQ3_K_M115details
Meta Llama 3.1 8B Instruct8.0BQ6_K7.8 GBQ3_K_M115details
Aya Expanse 8B8.0BQ6_K8.0 GBQ2_K115details
Ling 3.0 Tiny7.9B (1.4B act.)Q6_K7.3 GBQ6_K635details
MiMo 7B RL7.8BQ6_K7.6 GBQ2_K121details
InternLM2.5 7B Chat7.7BQ6_K7.5 GBQ3_K_M120details
InternLM2 Chat 7B7.7BQ6_K7.5 GBQ3_K_M120details
Qwen2.5 7B Instruct7.6BQ6_K7.0 GBQ6_K122details
DeepSeek R1 Distill Qwen 7B7.6BQ6_K7.0 GBQ6_K122details
Qwen2.5 Coder 7B Instruct7.6BQ6_K7.0 GBQ6_K122details
Olmo 3 7B Instruct7.3BQ5_K_M7.9 GBQ3_K_M146details
Olmo 3 1025 7B7.3BQ5_K_M7.9 GBQ3_K_M146details
Olmo 3 7B Think7.3BQ5_K_M7.9 GBQ3_K_M146details
Mistral 7B v0.17.2BQ6_K7.0 GBQ4_K_M128details
Mistral 7B Instruct v0.27.2BQ6_K7.0 GBQ4_K_M128details
Mistral 7B Instruct v0.17.2BQ6_K7.0 GBQ4_K_M128details
Falcon 7B7.2BQ8_07.7 GBQ8_099details
Falcon 7B Instruct7.2BQ8_07.4 GBQ8_0103details
Granite 4.0 H Tiny6.9B (1.5B act.)Q8_07.5 GBQ8_0487details
OLMoE 1B 7B 09246.9B (1.3B act.)Q8_08.4 GBQ4_K_M557details
DeepSeek Coder 7B Instruct v1.56.9BQ4_K_M8.3 GBno180details
DeepSeek Coder 6.7B Instruct6.7BQ4_K_M8.3 GBno186details
DeepSeek Coder 6.7B Base6.7BQ4_K_M8.3 GBno186details
Llama 2 7B HF6.7BQ4_K_M8.3 GBno185details
Yi 6B 200K6.1BQ8_07.0 GBQ8_0118details
Yi 6B6.1BQ8_07.0 GBQ8_0118details
Molmo2 4B4.9BQ8_06.2 GBQ5_K_M162details
WeVisOne VL 4B4.8BQ8_06.2 GBQ5_K_M162details
Mage VL Cowork4.7BQ8_06.2 GBQ5_K_M162details
Qianfan OCR4.7BQ8_06.2 GBQ5_K_M162details
Qwen3.5 4B4.7BQ8_05.4 GBQ8_0164details
SingVL S1 4B4.5BQ8_05.3 GBQ8_0170details
LightOnOCR 3 4B4.5BQ8_05.3 GBQ8_0170details
Intern Decision 4B4.5BQ8_05.3 GBQ8_0170details
Qwen3 VL 4B Instruct4.4BQ8_05.8 GBQ5_K_M178details
Gemma 3 4B IT4.3BF168.4 GBQ8_098details
MedGemma 1.5 4B IT4.3BF168.4 GBQ8_098details
MedGemma 4B IT4.3BF168.4 GBQ8_098details
TranslateGemma 4B IT4.3BQ8_05.0 GBQ8_0184details
Gemma 3 4B PT4.3BF168.4 GBQ8_098details
Phi 3.5 Vision Instruct4.1BQ8_07.6 GBno173details
Qwen3 4B4.0BQ8_05.8 GBQ6_K178details
Qwen3 4B Instruct 25074.0BQ8_05.8 GBQ6_K178details
Granite Vision 4.1 4B4.0BF167.6 GBQ8_0112details
Phi 4 Mini Instruct3.8BQ8_05.6 GBQ6_K186details
Phi 3 Mini 4K Instruct3.8BQ8_07.2 GBno187details
Phi 3.5 Mini Instruct3.8BQ8_07.2 GBno187details
Phi 3 Mini 128K Instruct3.8BQ8_07.2 GBno187details
Qwen2.5 VL 3B Instruct3.8BF168.0 GBQ8_0101details
Granite 4.2 3B3.7BF168.0 GBQ8_0104details
Granite 4.1 3B Llmon Preview3.7BF168.0 GBQ8_0104details
Raon 1 Nano3.5BF167.0 GBF16121details
Granite 4.1 3B3.4BF167.6 GBQ8_0112details
Granite 4.0 Micro3.4BF167.6 GBQ8_0112details
DeepSeek OCR 23.4B (1.2B act.)F166.6 GBF16367details
DeepSeek VL2 Tiny3.4B (1.2B act.)F167.4 GBQ8_0323details
Tiny Aya Global3.3BF167.5 GBF16113details
Unlimited OCR 23.3B (1.1B act.)F166.6 GBF16376details
Unlimited OCR3.3B (1.1B act.)F166.6 GBF16378details
DeepSeek OCR3.3B (1.1B act.)F166.6 GBF16378details
Granite 3.1 3B A800M Instruct3.3B (883M act.)F167.1 GBQ8_0430details
Llama 3.2 3B Instruct3.2BF167.5 GBQ8_0118details
Llama 3.2 3B3.2BF167.5 GBQ8_0118details
Hermes 3 Llama 3.2 3B3.2BF167.5 GBQ8_0118details
d1 3B3.1BF165.8 GBF16141details
LFM2.5 VL 3B3.1BF165.8 GBF16141details
Qwen2.5 3B Instruct3.1BF166.7 GBF16123details
SmolLM3 3B3.1BF166.9 GBQ8_0123details
SmolLM3 3B Base3.1BF166.9 GBQ8_0124details
Phi 22.8BF168.2 GBno137details
LFM2.5 2.6B2.7BQ8_03.4 GBQ8_0264details
Granite 3.3 2B Instruct2.5BF165.8 GBF16150details
MiniCPM5 2B2.5BF165.7 GBF16151details
WeVisOne VL 2B2.4BF165.4 GBF16186details
Cosmos Reason2 2B2.4BF165.4 GBF16187details
Qwen3.5 2B2.3BF164.6 GBF16195details
SmolVLM Instruct2.2BF165.4 GBQ8_0210details
Qwen3 1.7B2.0BF165.4 GBF16187details
d1 2B1.9BF164.0 GBF16219details
LFM3 VL 2B1.9BF164.0 GBF16219details
DeepSeek R1 Distill Qwen 1.5B1.8BF164.2 GBF16213details
LFM3 1.7B1.7BF167.2 GBF16110details
LFM3 1.7B Base1.7BF164.0 GBF16219details
SmolLM2 1.7B Instruct1.7BF165.2 GBQ8_0222details
SmolLM2 1.7B1.7BF165.2 GBQ8_0222details
SmolLM 1.7B1.7BF165.2 GBQ8_0222details
Granite 4.0 1B1.6BF164.3 GBF16233details
Qwen2.5 1.5B Instruct1.5BF163.8 GBF16246details
OLMo 2 0425 1B1.5BF164.4 GBF16256details
DeepSeek Coder 1.3B Instruct1.3BF164.5 GBQ8_0282details
Llama 3.2 1B1.2BF163.2 GBF16307details
Llama 3.2 1B Instruct1.2BF163.2 GBF16307details
LFM2.5 1.2B Instruct1.2BF165.0 GBF16162details
LFM2 1.2B1.2BF163.0 GBF16324details
HunyuanOCR1.1BF163.5 GBF16339details
PaddleOCR VL 1.6959MF162.5 GBF16396details
Qwen3.5 0.8B873MF162.4 GBF16488details
LightOnOCR 3 0.8B853MF162.4 GBF16501details
FastVLM 0.5B759MF162.2 GBF16501details
Qwen3 0.6B752MF163.0 GBF16504details
GOT OCR2.0716MF162.8 GBF16531details
Nemotron 3.5 Parse631MF161.9 GBF16581details
GOT OCR 2.0 HF561MF162.5 GBF16678details
SmolVLM2 500M Video Instruct507MF161.6 GBF16926details
SmolVLM 500M Instruct507MF161.6 GBF16926details
Qwen2.5 0.5B Instruct494MF161.7 GBF16769details
LFM2.5 VL 450M449MF161.3 GBF161068details
SmolLM2 360M Instruct362MF161.5 GBF161047details
SmolLM2 360M362MF161.5 GBF161047details
LFM2.5 350M354MF161.9 GBF16534details

Bigger models with CPU offload

These popular models don't fit in the RTX 3080 10GB's VRAM at 4-bit, but run with llama.cpp offload, keeping part of the model in system RAM. Speed range is DDR4-3200 to DDR5-6000 (estimate). Method.

ModelParamsOffload plan
GLM 5.3 Flash321B MoEOffload Q2_K, 128 GB RAM · ~6.9–14 tok/s
Gemma 4 31B IT31BOffload Q4_K_M, 32 GB RAM · ~1.9–3.8 tok/s
gpt-oss 120B117B MoEOffload Q4_K_M, 64 GB RAM · ~13–26 tok/s
DeepSeek V4 Flash 0731304B MoEOffload Q2_K, 128 GB RAM · ~7.9–15 tok/s
DeepSeek V4 Flash291B MoEOffload Q2_K, 128 GB RAM · ~9.2–18 tok/s
GLM 4.5 Air110B MoEOffload Q4_K_M, 128 GB RAM · ~5.2–10 tok/s
MiniMax M2.5229B MoEOffload Q3_K_M, 128 GB RAM · ~6.7–14 tok/s
Hy3299B MoEOffload Q2_K, 128 GB RAM · ~5.3–11 tok/s
Nex N2.6321B MoEOffload Q2_K, 128 GB RAM · ~6.6–13 tok/s

Comparing cards? The RTX 3080 10GB (launch MSRP $699) is in best GPUs for local LLMs under $800. See also Mac vs PC.

Just too big at 4-bit

These need up to 35% more memory than the RTX 3080 10GB has at Q4_K_M and 8K context. A smaller quant or shorter context may work.

ModelQ4_K_M needsShort by
DeepSeek V2 Lite9.8 GB0.3 GB
LLaDA2.1 Mini10.3 GB0.8 GB
Kimi VL A3B Thinking10.3 GB0.8 GB
Qwen3 14B10.3 GB0.8 GB
DeepSeek Coder V2 Lite Instruct10.5 GB1.0 GB
DeepSeek V2 Lite Chat10.5 GB1.0 GB
DeepSeek R1 Distill Qwen 14B10.6 GB1.1 GB
Qwen2.5 Coder 14B Instruct10.6 GB1.1 GB
Phi 410.6 GB1.1 GB
Moonlight 16B A3B Instruct10.8 GB1.3 GB
Kimi VL A3B Thinking 250610.8 GB1.3 GB
Kimi VL A3B Instruct10.8 GB1.3 GB
gpt-oss Safeguard 20B11.8 GB2.3 GB
gpt-oss 20B11.8 GB2.3 GB
DeepSeek MoE 16B Base12.5 GB3.0 GB

Questions

What is the biggest model that runs on the RTX 3080 10GB?

Of the models we track, the largest that fits at 8K context is LFM2.5 24B A2B (24B) at Q2_K, needing about 9.2 GB.

How much of the RTX 3080 10GB's memory can a model use?

About 9.5 GB: we assume 95% of the 10 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.