Skip to content

What can the Arc B580 run?

What fits on the Arc B580
Text models: 246Visual models: 12Strongest first
Memory
12 GB
Bandwidth
456 GB/s
Maker
Intel

Open the calculator on this machineEvery model for 12 GB of VRAM

Text models that run on it

Text models that run on the Arc B580, best first
ModelBest fitContextSpeedStrongest at
Qwen3.5 9B8-bit24K24 tok/sChat, Coding, Agents, Docs
Gemma 4 12B6-bit112K22 tok/sChat, Coding, Docs
Qwen3.5 4B8-bit176K42 tok/sAgents, Docs
Ministral 3 14B4-bit16K27 tok/sCoding, Agents, Docs
Gemma 4 E4B8-bit128K37 tok/sChat, Coding, Docs
gpt-oss 20B3-bit40K99 tok/sAgents
Ministral 3 8B6-bit28K31 tok/sDocs
Gemma 4 E2B8-bit128K67 tok/sChat
Nemotron 3 Nano 4B8-bit256K51 tok/sDocs
Llama 3.1 8B8-bit20K27 tok/sChat, Coding, Agents, Docs
Qwen3 0.6B8-bit40K193 tok/sNot rated
Qwen3 14B3-bit24K31 tok/sNot rated
Qwen3 VL 8B Instruct6-bit24K32 tok/sText and images
Qwen3 8B8-bit16K27 tok/sNot rated
Qwen2.5 7B Instruct8-bit32K29 tok/sNot rated
Qwen2.5 VL 7B Instruct8-bit40K26 tok/sText and images
Unlimited OCR8-bit32K153 tok/sReading documents
Qwen2.5 1.5B Instruct8-bit32K113 tok/sNot rated
Qwen3.5 2B8-bit256K82 tok/sText and images
Qwen3 VL 4B Instruct8-bit40K44 tok/sText and images
Qwen3 1.7B8-bit40K90 tok/sNot rated
Qwen3 4B8-bit40K48 tok/sNot rated
Qwen3 4B Instruct 25078-bit44K48 tok/sNot rated
Qwen2.5 3B Instruct8-bit32K60 tok/sNot rated
DeepSeek OCR8-bit8K153 tok/sReading documents
Qwen2.5 Coder 14B Instruct3-bit20K30 tok/sCoding
Qwen2.5 VL 3B Instruct8-bit125K51 tok/sText and images
Qwen2.5 Coder 7B Instruct8-bit32K29 tok/sCoding
Mistral 7B Instruct V0.28-bit28K29 tok/sNot rated
Chandra OCR 28-bit160K37 tok/sReading documents
Qwen3 VL 2B Instruct8-bit72K86 tok/sText and images
Qwen2 VL 2B Instruct8-bit32K84 tok/sText and images
MiniCPM5 2B8-bit128K77 tok/sNot rated
Gemma 3 4B IT8-bit128K53 tok/sText and images
Llama 3.2 3B Instruct8-bit64K60 tok/sNot rated
LightOnOCR 2 1B8-bit16K157 tok/sText and images
Qwen2.5 14B Instruct3-bit20K30 tok/sNot rated
Qwen3 1.7B Base8-bit32K103 tok/sNot rated
HunyuanOCR8-bit96K150 tok/sText and images
Qwen3 0.6B Base8-bit32K224 tok/sNot rated
Rax 4.58-bit256K82 tok/sText and images
Qwen2 VL 7B Instruct8-bit32K26 tok/sText and images
Ornith 1.5 9B8-bit20K24 tok/sNot rated
Unlimited OCR 28-bit120K153 tok/sReading documents
Granite 4.1 3B8-bit88K55 tok/sNot rated
DeepSeek R1 Distill Qwen 1.5B8-bit128K113 tok/sNot rated
Mistral 7B V0.18-bit32K29 tok/sNot rated
Qwen2.5 Coder 3B Instruct8-bit32K60 tok/sCoding
Meta Llama 3 8B Instruct Abliterated V38-bit8K27 tok/sNot rated
EXAONE 3.5 7.8B Instruct8-bit20K27 tok/sNot rated
Saiga Llama3 8B8-bit8K27 tok/sNot rated
Qwen2.5 Coder 1.5B Instruct8-bit32K113 tok/sCoding
DeepSeek R1 0528 Qwen3 8B8-bit16K27 tok/sNot rated
Gemma 3 12B IT5-bit28K26 tok/sText and images
Phi 3.5 Mini Instruct8-bit16K52 tok/sNot rated
Qwen3 4B Thinking 25078-bit44K48 tok/sNot rated
Phi 4 Mini Instruct8-bit52K51 tok/sNot rated
Dots.ocr8-bit128K68 tok/sReading documents
DeepSeek R1 Distill Qwen 7B8-bit52K29 tok/sNot rated
Phi 3 Mini 4K Instruct8-bit4K52 tok/sNot rated

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 246 are shown.

Image and video models

Image and video models that run on the Arc B580
ModelBest fitPer outputStrongest at
Stable Diffusion XL 1.016-bit14 sImages
Z-Image Turbo8-bit19 sImages
FLUX.1 [dev]6-bit43 sImages
Qwen-Image 2.18-bit5.1 minImages
FLUX.3 Klein16-bit28 sImages
FLUX.2 [klein] 9B8-bit9.4 sImages
Krea 2 Turbo5-bit43 sImages
Stable Diffusion 3.5 Large8-bit2.7 minImages
MOSS-Image16-bit3.3 minImages
Wan 2.2 TI2V-5B16-bit29 minVideo
MOSS Image Base16-bit3.3 minImages
HunyuanVideo 1.56-bit2.4 hVideo

Only just too big

Questions

What is the best LLM for the Arc B580?

Of the open models we track, the strongest that run well on the Arc B580 are Qwen3.5 9B, Gemma 4 12B, Qwen3.5 4B. 246 text models fit it in all.

Calculated from published specs, not benchmarked. All machines