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

What AI models run on the Arc B580?

The Arc B580 has 12 GB of VRAM (456 GB/s). About 11.4 GB is usable for a model. 173 of 298 open models we track fit at 8K context.

173models fit (8K ctx)
154still fit at 32K ctx
31Blargest that fits
456 GB/smemory bandwidth

This card has 12 GB of VRAM. See every model that fits in 12 GB VRAM. Other 12 GB hardware: RTX 3060 12GB, RTX 3080 12GB, RTX 4070, RTX 4070 Super, RTX 4070 Ti, RTX 5070, RX 7700 XT.

Compare for local LLMs: Arc B580 vs RTX 4060Arc B580 vs RTX 3060 12GB all comparisons

Using a specific tool? Ollama models that run on the Arc B580 · LM Studio · llama.cpp offload

Models that fit, largest first

ModelParamsBest quant (8K)Needs32K contextMax tok/s (est.)Details
GLM 4.7 Flash31B (4.6B act.)Q2_K11.4 GBno280details
Muse Glimmer 30B30BQ2_K11.4 GBno41details
Raon 1 Mini Preview29B (4.3B act.)Q2_K11.2 GBno282details
Arete 27B27BQ2_K11.2 GBno43details
MiniMax H3 Image PE27BQ2_K11.3 GBno43details
Trinity Mini26B (2.0B act.)Q2_K9.8 GBQ2_K643details
GLiDE 27B26BQ2_K10.5 GBno47details
DiffusionGemma 26B A4B IT26B (4.0B act.)Q2_K10.4 GBQ2_K304details
LFM2.5 24B A2B24B (2.0B act.)Q2_K9.2 GBQ2_K608details
ERNIE 4.5 21B A3B PT22BQ2_K8.6 GBQ2_K56details
ERNIE 4.5 21B A3B Thinking22BQ2_K8.6 GBQ2_K56details
Kimi VL A3B Thinking 250616B (3.0B act.)Q3_K_M8.4 GBQ3_K_M311details
Kimi VL A3B Thinking16B (3.0B act.)Q3_K_M8.4 GBQ3_K_M311details
Kimi VL A3B Instruct16B (3.0B act.)Q3_K_M8.7 GBQ3_K_M301details
DeepSeek MoE 16B Base16B (2.8B act.)Q3_K_M10.2 GBno311details
LLaDA2.1 Mini16B (1.4B act.)Q3_K_M8.4 GBQ3_K_M651details
Moonlight 16B A3B Instruct16B (3.0B act.)Q3_K_M8.7 GBQ3_K_M293details
DeepSeek Coder V2 Lite Instruct16B (2.7B act.)Q3_K_M8.0 GBQ3_K_M351details
DeepSeek V2 Lite16B (2.7B act.)Q4_K_M9.8 GBQ3_K_M280details
DeepSeek V2 Lite Chat16B (2.7B act.)Q3_K_M8.4 GBQ3_K_M331details
Qwen2.5 Coder 14B Instruct15BQ3_K_M9.0 GBno62details
DeepSeek R1 Distill Qwen 14B15BQ3_K_M9.0 GBno62details
Qwen3 14B15BQ3_K_M8.7 GBQ2_K63details
Phi 415BQ3_K_M9.0 GBno62details
Gemma 3 12B IT12BQ5_K_M9.6 GBQ4_K_M54details
Llama Guard 4 12B12BQ5_K_M10.2 GBQ2_K53details
KORMo Web 11B11BQ5_K_M9.3 GBQ2_K57details
Llama 3.2 11B Vision Instruct11BQ6_K10.0 GBQ2_K52details
Llama 3.2 11B Vision11BQ6_K10.0 GBQ2_K52details
GLM 4.6V Flash10BQ6_K8.7 GBQ6_K55details
GLM 4.1V 9B Thinking10BQ6_K8.7 GBQ6_K55details
Step3 VL 10B10BQ8_09.9 GBQ4_K_M52details
SenseNova Loop 9B Preview9.7BQ8_010.0 GBQ6_K48details
Qwen3.5 9B9.7BQ8_010.3 GBQ6_K47details
Ornith 1.5 9B9.7BQ8_010.0 GBQ6_K48details
SenseNova Loop 9B9.4BQ8_010.0 GBQ6_K48details
MiMo V2.6 Distill Qwen 9B9.4BQ8_010.0 GBQ6_K48details
GLM 4 9B9.4BQ6_K8.7 GBQ6_K55details
Apertus v1.5 8B8.9BQ6_K8.4 GBQ4_K_M62details
InternLM3 8B Instruct8.8BQ8_09.7 GBQ6_K49details
Granite 4.2 8B8.8BQ6_K8.6 GBQ3_K_M63details
Granite 4.1 8B8.8BQ6_K8.6 GBQ3_K_M63details
Granite 4.1 8B Llmon Preview8.8BQ6_K8.6 GBQ3_K_M63details
Qwen3 VL 8B Instruct8.8BQ8_09.9 GBQ4_K_M52details
Molmo2 8B8.7BQ8_09.9 GBQ4_K_M52details
LFM2.5 8B A1B8.5B (1.0B act.)Q8_09.1 GBQ8_0429details
Qwen2.5 VL 7B Instruct8.3BQ8_08.7 GBQ8_056details
UI TARS 1.5 7B8.3BQ8_08.7 GBQ8_056details
olmOCR 2 7B 10258.3BQ8_08.7 GBQ8_056details
olmOCR 7B 0225 Preview8.3BQ8_08.7 GBQ8_056details
UI TARS 7B DPO8.3BQ8_08.7 GBQ8_056details
Qwen3 8B8.2BQ8_09.9 GBQ4_K_M52details
DeepSeek R1 0528 Qwen3 8B8.2BQ8_09.9 GBQ4_K_M52details
Granite 3.3 8B Instruct8.2BQ8_09.8 GBQ4_K_M53details
Granite 3.1 8B Instruct8.2BQ8_09.8 GBQ4_K_M53details
Granite 3.2 8B Instruct8.2BQ8_09.8 GBQ4_K_M53details
Apertus 8B Instruct 25098.1BQ8_09.6 GBQ5_K_M53details
Llama 3.1 8B Instruct8.0BQ8_09.6 GBQ5_K_M53details
Llama 3.1 8B8.0BQ8_09.6 GBQ5_K_M53details
Hermes 3 Llama 3.1 8B8.0BQ8_09.6 GBQ5_K_M53details
Llama Guard 3 8B8.0BQ8_09.6 GBQ5_K_M53details
DeepSeek R1 Distill Llama 8B8.0BQ8_09.6 GBQ5_K_M53details
Meta Llama 3 8B Instruct8.0BQ8_09.6 GBQ5_K_M53details
Meta Llama Guard 2 8B8.0BQ8_09.6 GBQ5_K_M53details
Meta Llama 3.1 8B Instruct8.0BQ8_09.6 GBQ5_K_M53details
Aya Expanse 8B8.0BQ8_09.8 GBQ4_K_M53details
Ling 3.0 Tiny7.9B (1.4B act.)Q8_08.7 GBQ8_0310details
MiMo 7B RL7.8BQ8_09.4 GBQ4_K_M56details
InternLM2.5 7B Chat7.7BQ8_09.2 GBQ5_K_M55details
InternLM2 Chat 7B7.7BQ8_09.2 GBQ5_K_M55details
Qwen2.5 7B Instruct7.6BQ8_08.7 GBQ8_056details
DeepSeek R1 Distill Qwen 7B7.6BQ8_08.7 GBQ8_056details
Qwen2.5 Coder 7B Instruct7.6BQ8_08.7 GBQ8_056details
Olmo 3 7B Instruct7.3BQ6_K8.7 GBQ4_K_M76details
Olmo 3 1025 7B7.3BQ6_K8.7 GBQ4_K_M76details
Olmo 3 7B Think7.3BQ6_K8.7 GBQ4_K_M76details
Mistral 7B v0.17.2BQ8_08.6 GBQ6_K59details
Mistral 7B Instruct v0.27.2BQ8_08.6 GBQ6_K59details
Mistral 7B Instruct v0.17.2BQ8_08.6 GBQ6_K59details
Falcon 7B7.2BQ8_07.7 GBQ8_059details
Falcon 7B Instruct7.2BQ8_07.4 GBQ8_062details
Granite 4.0 H Tiny6.9B (1.5B act.)Q8_07.5 GBQ8_0292details
OLMoE 1B 7B 09246.9B (1.3B act.)Q8_08.4 GBQ6_K334details
DeepSeek Coder 7B Instruct v1.56.9BQ6_K9.6 GBno80details
DeepSeek Coder 6.7B Instruct6.7BQ6_K9.6 GBno82details
DeepSeek Coder 6.7B Base6.7BQ6_K9.6 GBno82details
Llama 2 7B HF6.7BQ6_K9.6 GBno83details
Yi 6B 200K6.1BQ8_07.0 GBQ8_071details
Yi 6B6.1BQ8_07.0 GBQ8_071details
Molmo2 4B4.9BF1610.0 GBQ8_052details
WeVisOne VL 4B4.8BF1610.0 GBQ8_052details
Mage VL Cowork4.7BF1610.0 GBQ8_052details
Qianfan OCR4.7BF1610.1 GBQ8_052details
Qwen3.5 4B4.7BF169.2 GBF1653details
SingVL S1 4B4.5BF169.0 GBF1654details
LightOnOCR 3 4B4.5BF169.0 GBF1654details
Intern Decision 4B4.5BF169.0 GBF1654details
Qwen3 VL 4B Instruct4.4BF169.3 GBQ8_057details
Gemma 3 4B IT4.3BF168.4 GBF1659details
MedGemma 1.5 4B IT4.3BF168.4 GBF1659details
MedGemma 4B IT4.3BF168.4 GBF1659details
TranslateGemma 4B IT4.3BF169.2 GBF1653details
Gemma 3 4B PT4.3BF168.4 GBF1659details
Phi 3.5 Vision Instruct4.1BQ8_07.6 GBno104details
Qwen3 4B4.0BF169.3 GBQ8_057details
Qwen3 4B Instruct 25074.0BF169.3 GBQ8_057details
Granite Vision 4.1 4B4.0BF167.6 GBF1667details
Phi 4 Mini Instruct3.8BF168.9 GBQ8_059details
Phi 3 Mini 4K Instruct3.8BQ8_07.2 GBno112details
Phi 3.5 Mini Instruct3.8BQ8_07.2 GBno112details
Phi 3 Mini 128K Instruct3.8BQ8_07.2 GBno112details
Qwen2.5 VL 3B Instruct3.8BF168.0 GBF1661details
Granite 4.2 3B3.7BF168.0 GBF1662details
Granite 4.1 3B Llmon Preview3.7BF168.0 GBF1662details
Raon 1 Nano3.5BF167.0 GBF1673details
Granite 4.1 3B3.4BF167.6 GBF1667details
Granite 4.0 Micro3.4BF167.6 GBF1667details
DeepSeek OCR 23.4B (1.2B act.)F166.6 GBF16220details
DeepSeek VL2 Tiny3.4B (1.2B act.)F167.4 GBF16194details
Tiny Aya Global3.3BF167.5 GBF1668details
Unlimited OCR 23.3B (1.1B act.)F166.6 GBF16226details
Unlimited OCR3.3B (1.1B act.)F166.6 GBF16227details
DeepSeek OCR3.3B (1.1B act.)F166.6 GBF16227details
Granite 3.1 3B A800M Instruct3.3B (883M act.)F167.1 GBF16258details
Llama 3.2 3B Instruct3.2BF167.5 GBF1671details
Llama 3.2 3B3.2BF167.5 GBF1671details
Hermes 3 Llama 3.2 3B3.2BF167.5 GBF1671details
d1 3B3.1BF165.8 GBF1684details
LFM2.5 VL 3B3.1BF165.8 GBF1684details
Qwen2.5 3B Instruct3.1BF166.7 GBF1674details
SmolLM3 3B3.1BF166.9 GBF1674details
SmolLM3 3B Base3.1BF166.9 GBF1674details
Phi 22.8BF168.2 GBno82details
LFM2.5 2.6B2.7BQ8_03.4 GBQ8_0159details
Granite 3.3 2B Instruct2.5BF165.8 GBF1690details
MiniCPM5 2B2.5BF165.7 GBF1690details
WeVisOne VL 2B2.4BF165.4 GBF16112details
Cosmos Reason2 2B2.4BF165.4 GBF16112details
Qwen3.5 2B2.3BF164.6 GBF16117details
SmolVLM Instruct2.2BF165.4 GBF16126details
Qwen3 1.7B2.0BF165.4 GBF16112details
d1 2B1.9BF164.0 GBF16132details
LFM3 VL 2B1.9BF164.0 GBF16132details
DeepSeek R1 Distill Qwen 1.5B1.8BF164.2 GBF16128details
LFM3 1.7B1.7BF167.2 GBF1666details
LFM3 1.7B Base1.7BF164.0 GBF16132details
SmolLM2 1.7B Instruct1.7BF165.2 GBF16133details
SmolLM2 1.7B1.7BF165.2 GBF16133details
SmolLM 1.7B1.7BF165.2 GBF16133details
Granite 4.0 1B1.6BF164.3 GBF16140details
Qwen2.5 1.5B Instruct1.5BF163.8 GBF16148details
OLMo 2 0425 1B1.5BF164.4 GBF16153details
DeepSeek Coder 1.3B Instruct1.3BF164.5 GBF16169details
Llama 3.2 1B1.2BF163.2 GBF16184details
Llama 3.2 1B Instruct1.2BF163.2 GBF16184details
LFM2.5 1.2B Instruct1.2BF165.0 GBF1697details
LFM2 1.2B1.2BF163.0 GBF16195details
HunyuanOCR1.1BF163.5 GBF16204details
PaddleOCR VL 1.6959MF162.5 GBF16238details
Qwen3.5 0.8B873MF162.4 GBF16293details
LightOnOCR 3 0.8B853MF162.4 GBF16301details
FastVLM 0.5B759MF162.2 GBF16300details
Qwen3 0.6B752MF163.0 GBF16302details
GOT OCR2.0716MF162.8 GBF16318details
Nemotron 3.5 Parse631MF161.9 GBF16349details
GOT OCR 2.0 HF561MF162.5 GBF16407details
SmolVLM2 500M Video Instruct507MF161.6 GBF16556details
SmolVLM 500M Instruct507MF161.6 GBF16556details
Qwen2.5 0.5B Instruct494MF161.7 GBF16462details
LFM2.5 VL 450M449MF161.3 GBF16641details
SmolLM2 360M Instruct362MF161.5 GBF16628details
SmolLM2 360M362MF161.5 GBF16628details
LFM2.5 350M354MF161.9 GBF16320details

Bigger models with CPU offload

These popular models don't fit in the Arc B580'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 · ~6.8–13 tok/s
gpt-oss 120B117B MoEOffload Q4_K_M, 64 GB RAM · ~13–25 tok/s
DeepSeek V4 Flash 0731304B MoEOffload Q2_K, 128 GB RAM · ~7.3–13 tok/s
DeepSeek V4 Flash291B MoEOffload Q2_K, 128 GB RAM · ~9.1–17 tok/s
GLM 4.5 Air110B MoEOffload Q4_K_M, 128 GB RAM · ~5.1–9.7 tok/s
MiniMax M2.5229B MoEOffload Q3_K_M, 128 GB RAM · ~6.7–13 tok/s
Hy3299B MoEOffload Q3_K_M, 128 GB RAM · ~4.0–7.9 tok/s

Comparing cards? The Arc B580 (launch MSRP $249) is in best GPUs for local LLMs under $300. See also Mac vs PC.

Just too big at 4-bit

These need up to 35% more memory than the Arc B580 has at Q4_K_M and 8K context. A smaller quant or shorter context may work.

ModelQ4_K_M needsShort by
gpt-oss Safeguard 20B11.8 GB0.4 GB
gpt-oss 20B11.8 GB0.4 GB
DeepSeek MoE 16B Base12.5 GB1.1 GB
ERNIE 4.5 21B A3B PT13.6 GB2.2 GB
ERNIE 4.5 21B A3B Thinking13.6 GB2.2 GB
LFM2.5 24B A2B14.3 GB2.9 GB

Questions

What is the biggest model that runs on the Arc B580?

Of the models we track, the largest that fits at 8K context is GLM 4.7 Flash (31B) at Q2_K, needing about 11.4 GB.

How much of the Arc B580's memory can a model use?

About 11.4 GB: we assume 95% of the 12 GB is usable after driver and display overhead.

Spec source: Intel specifications. Model data: Hugging Face Hub API. Speed is a bandwidth ceiling (bandwidth ÷ bytes read per token), not a benchmark. Methodology · Report an error.