Arc B580 12G — what LLMs can it run?
57 of 112 indexed models fit comfortably in 12GB at 4K context, each at the highest-quality quant that still leaves headroom.
The short answer for Arc B580 12G
12 GB, 456 GB/s GDDR6. 57 of 112 models in this index fit comfortably at 4K context; 58 load at all.
- Biggest that fits
- DeepSeek-Coder-V2-Lite Instruct — 10.1 GB at Q4_K_M — 1.9 GB spare at 4K, so longer context comes out of a thin margin.
- Room to grow
- Qwen2-VL 7B Instruct — 5.5 GB at Q4_K_M — under 60% of the card, which leaves 6.5 GB for a long context window or a second process.
- Speed ceiling
- Llama 3.1 8B Instruct at Q4_K_M reads 4.6 GB of weights per token, so 456 GB/s puts a hard ceiling near 99 tok/s. That is arithmetic on two published numbers, not a benchmark — real throughput lands below it. No run on this card has been measured here, so there is nothing to compare the ceiling against.
- The wall you will hit
- Backend coverage, not capacity. At 456 GB/s the card has the memory system for its size class; what limits it is which runtimes reach it. llama.cpp's SYCL backend lists Arc A-series and B580 cards as verified, and Ollama runs them through Vulkan — so GGUF is the format this site recommends here. AWQ, GPTQ and EXL2 are not.
14B · 18
| Model | Quant | Est. VRAM | Headroom | tok/s |
|---|---|---|---|---|
| DeepSeek-Coder-V2-Lite Instruct16B-A2.4B | Q4_K_M | 10.08 GB | +1.9 GB | ≤330Ceiling |
| DeepSeek-V2-Lite Chat16B-A2.4B | Q4_K_M | 10.08 GB | +1.9 GB | ≤330Ceiling |
| StarCoder2 15B15B | Q4_K_M | 10.19 GB | +1.8 GB | 29–39Derived |
| Qwen3 14B Instruct14B | Q4_K_M | 10.04 GB | +2 GB | 30–39Derived |
| Qwen2.5 14B Instruct14B | Q4_K_M | 10.14 GB | +1.9 GB | 30–39Derived |
| DeepSeek-R1-Distill-Qwen-14B14B | Q4_K_M | 10.14 GB | +1.9 GB | 30–39Derived |
| Phi-4 14B14B | Q4_K_M | 10.17 GB | +1.8 GB | 30–39Derived |
| Phi-3 Medium 14B Instruct14B | Q4_K_M | 9.73 GB | +2.3 GB | 31–41Derived |
| Mistral Nemo 12B Instruct12B | Q4_K_M | 8.42 GB | +3.6 GB | 36–48Derived |
| Gemma 3 12B IT12B | Q5_K_M | 9.84 GB | +2.2 GB | 31–41Derived |
| Mellum2.1 12B-A2.5B Thinking12B-A2.5B | Q4_K_M | 8.59 GB | +3.4 GB | ≤288Ceiling |
| Stable LM 2 12B Chat12B | Q4_K_M | 8.35 GB | +3.7 GB | 36–48Derived |
| Gemma 4 12B IT12B | Q4_K_M | 8.02 GB | +4 GB | 37–49Derived |
| Llama 3.2 11B Vision Instruct11B | Q4_K_M | 7.66 GB | +4.3 GB | 40–53Derived |
| Solar 10.7B Instruct11B | Q4_K_M | 7.6 GB | +4.4 GB | 41–54Derived |
| Falcon 3 10B Instruct10B | Q4_K_M | 7.28 GB | +4.7 GB | 42–56Derived |
| Gemma 2 9B Instruct9B | Q4_K_M | 7.3 GB | +4.7 GB | 47–63Derived |
| Qwen3.5 9B9B | Q8_0 | 10.37 GB | +1.6 GB | 27–36Derived |
7B · 25
| Model | Quant | Est. VRAM | Headroom | tok/s |
|---|---|---|---|---|
| GLM-4-9B-Chat9B | Q8_0 | 10.16 GB | +1.8 GB | 28–37Derived |
| Qwen3-VL 8B Instruct8B | Q8_0 | 10.39 GB | +1.6 GB | 28–38Derived |
| Granite 4.2 8B8B | Q8_0 | 10.45 GB | +1.6 GB | 28–38Derived |
| Ministral 3 8B Instruct8B | Q8_0 | 10.02 GB | +2 GB | 29–39Derived |
| Qwen2-VL 7B Instruct7B | Q4_K_M | 5.49 GB | +6.5 GB | 53–70Derived |
| Granite 3.1 8B Instruct8B | Q4_K_M | 5.88 GB | +6.1 GB | 53–71Derived |
| Qwen3 8B Instruct8B | Q6_K | 7.64 GB | +4.4 GB | 40–52Derived |
| Llama 3.1 8B Instruct8B | Q8_0 | 9.47 GB | +2.5 GB | 31–41Derived |
| Nous Hermes 3 Llama 3.1 8B8B | Q4_K_M | 5.64 GB | +6.4 GB | 55–72Derived |
| Aya 23 8B8B | Q4_K_M | 5.64 GB | +6.4 GB | 55–72Derived |
| OpenChat 3.6 8B8B | Q4_K_M | 5.64 GB | +6.4 GB | 55–72Derived |
| DeepSeek-R1-Distill-Llama-8B8B | Q5_K_M | 6.51 GB | +5.5 GB | 47–62Derived |
| InternLM2 7B Chat7B | Q4_K_M | 5.45 GB | +6.6 GB | 57–75Derived |
| Qwen2.5 7B Instruct7B | Q6_K | 6.77 GB | +5.2 GB | 42–56Derived |
| Qwen2.5-Coder 7B Instruct7B | Q4_K_M | 5.07 GB | +6.9 GB | 57–76Derived |
| DeepSeek-R1-Distill-Qwen-7B7B | Q4_K_M | 5.07 GB | +6.9 GB | 57–76Derived |
| Gemma 4 E4B ITE4B | Q8_0 | 8.46 GB | +3.5 GB | 33–44Derived |
| OLMo 2 7B Instruct7B | Q8_0 | 10.3 GB | +1.7 GB | 34–45Derived |
| OLMo 3 7B Instruct7B | Q8_0 | 10.3 GB | +1.7 GB | 34–45Derived |
| WizardLM-2 7B7B | Q4_K_M | 5.14 GB | +6.9 GB | 61–80Derived |
| Mistral 7B Instruct v0.37B | Q6_K | 6.75 GB | +5.3 GB | 45–59Derived |
| Zephyr 7B Beta7B | Q6_K | 6.75 GB | +5.3 GB | 45–59Derived |
| Gemma 4 E2B ITE2B | Q8_0 | 5.2 GB | +6.8 GB | ≤97Ceiling |
| Qwen3.5 4B4B | Q8_0 | 4.97 GB | +7 GB | ≤104Ceiling |
| Gemma 3 4B IT4B | Q8_0 | 5.05 GB | +7 GB | ≤105Ceiling |
≤3B · 14
| Model | Quant | Est. VRAM | Headroom | tok/s |
|---|---|---|---|---|
| Qwen3 4B Instruct4B | Q6_K | 4.05 GB | +8 GB | ≤146Ceiling |
| Phi-4 Mini Instruct3.8B | Q8_0 | 4.81 GB | +7.2 GB | ≤118Ceiling |
| Phi-3.5 Mini Instruct3.8B | Q8_0 | 5.89 GB | +6.1 GB | ≤118Ceiling |
| Granite 4.2 3B3B | Q8_0 | 4.41 GB | +7.6 GB | ≤124Ceiling |
| Llama 3.2 3B Instruct3B | Q8_0 | 4.05 GB | +8 GB | ≤141Ceiling |
| Qwen2.5 3B Instruct3B | Q8_0 | 3.59 GB | +8.4 GB | ≤146Ceiling |
| SmolLM3 3B3B | Q8_0 | 3.73 GB | +8.3 GB | ≤147Ceiling |
| Gemma 2 2B Instruct2B | Q8_0 | 3.34 GB | +8.7 GB | ≤173Ceiling |
| Qwen3.5 2B2B | Q8_0 | 2.23 GB | +9.8 GB | ≤230Ceiling |
| Qwen3 1.7B Instruct1.7B | Q8_0 | 2.39 GB | +9.6 GB | ≤262Ceiling |
| Qwen2.5 1.5B Instruct1.5B | Q8_0 | 1.83 GB | +10.2 GB | ≤294Ceiling |
| Llama 3.2 1B Instruct1B | Q8_0 | 1.51 GB | +10.5 GB | ≤365Ceiling |
| Qwen3.5 0.8B0.8B | Q8_0 | 0.94 GB | +11.1 GB | ≤563Ceiling |
| Qwen2.5 0.5B Instruct0.5B | Q8_0 | 0.6 GB | +11.4 GB | ≤931Ceiling |
How this list is built
Each row is the lowest-perplexity-loss quant of that model whose estimated total — weights plus KV cache at 4K context plus activation buffer — uses at most 88% of the card. That is the calculator's "green" threshold, so every row here has real headroom rather than only just fitting. Raise the context length and the list shortens; the calculator lets you check any combination directly. The speed column is this card: a measured run, a range scaled from this card's own Llama 2 7B benchmark, or a bandwidth ceiling. A figure that only exists for the RTX 4090 is labelled as that card.
A further 1 models load but with no headroom to spare (up to 105% of VRAM) — the Quant Hub’s GPU chips count those too, which is why its number is higher.
Measured on this card
No benchmark runs in this index were recorded on a Arc B580 12G. Every figure on this page is calculated from the model architecture and the quant level — treat them as estimates, not measurements.
Setup guides for this card
Step-by-step guides written for the Arc B580 12G's runtime or memory size, with the commands and the checks that prove the model is running on the card.
Stepping up
A Arc A770 16G (16GB) fits 3 more of the indexed models than this card. Arc A770 16G →
Before buying, hear the speed difference on a model both cards hold: Arc B580 12G vs Arc A770 16G (Qwen3 14B Instruct) →
Common questions
What is the best local LLM for an Arc B580 12G?
For everyday use, Qwen2-VL 7B Instruct at Q4_K_M — about 5.5 GB of the card's 12 GB at 4K context, leaving 6.5 GB for a longer window. If you want the largest thing that will load, that is DeepSeek-Coder-V2-Lite Instruct at Q4_K_M (10.1 GB, 1.9 GB spare). "Best" here means best fit for the memory budget — this index does not run task benchmarks, so it cannot tell you which model is smarter.
Can an Arc B580 12G run InternLM2 20B Chat?
No. Its smallest build here, Q4_K_M, needs about 13.4 GB and this card has 12.0 GB usable — short by 1.4 GB before any context beyond 4K. The largest model this card does clear is DeepSeek-Coder-V2-Lite Instruct.
How many tokens per second does an Arc B580 12G do on an 8B model at Q4?
This index has no measured run on this card, so it does not publish a figure. What can be stated from specifications: generating a token requires reading every weight once, Llama 3.1 8B Instruct at Q4_K_M is 4.6 GB of weights, and this card moves 456 GB/s — a ceiling near 99 tok/s. Batch size, context length, the runtime and how much of the model sits in cache all take you below it.
Arc B580 12G or RTX 3080 Ti for local LLMs?
They hold the same models: 12 GB against 12 GB fits 57 and 57 of 112 respectively at 4K. The difference is throughput — 456 GB/s against 912 GB/s, a 2.0× gap in how fast the weights can be read, which is what token generation is bound by. The RTX 3080 Ti generates faster on any model both can hold.
57 of 112 indexed models fit comfortably in 12GB at 4K context, each at the highest-quality quant that still leaves headroom.
This page's figures change when the model or the runtime does.Last updated 2026-10-20 RSS → /feed.xml