Skip to content
Run LLM Locally

Llama 3.2 3B Instruct on the Arc B580

With 12 GB of VRAM, the Arc B580 runs Llama 3.2 3B Instruct at Q4_K_M entirely on the GPU with an 8K context (about 3.8 GB needed). At Q8_0 it fits entirely.

ShareFacebookXWhatsAppLinkedInEmail

Llama 3.2 3B Instruct

Q4_K_M · 8K context

Excellent

The model needs about 3.8 GB and fits in the 11.6 GB available, with 7.8 GB to spare.

Your hardware
Arc B580 12 GB · 64 GB RAM
Estimated memory
~3.8 GB estimate
VRAM (usable memory)
11.6 GB
GPU only
Yes
CPU/GPU offload
Not required
Generation speed
69–108 tok/s estimate
Largest context in fast memory
76K tokens
Confidence: low

Based on: memory bandwidth (456 GB/s) and model size

Memory breakdown

~3.8 GB (3.5 GB–4.2 GB) needed · estimate

Fits entirely in VRAM · 7.8 GB left

  • Model weights1.9 GB
  • KV cache0.9 GB
  • Runtime overhead0.8 GB
  • Safety margin0.3 GB
Assumptions: context 8K tokens · KV cache f16 · llama.cpp (SYCL) · flash attention on · weights = real file size

EstimateData updated 24 days agoSource: Run LLM Locally engine, Hugging Face config.json + model card

What makes this estimate less reliable

  • no benchmark for this hardware and model yet

Other quantizations on the Arc B580

Llama 3.2 3B Instruct on Arc B580 + 64 GB RAM · 8K context
Llama 3.2 3B Instruct Q8_08K~5.2 GB · 44–69 tok/s estimateExcellent
Llama 3.2 3B Instruct Q6_K8K~4.4 GB · 55–86 tok/s estimateExcellent
Llama 3.2 3B Instruct Q5_K_M8K~4.1 GB · 61–96 tok/s estimateExcellent
Llama 3.2 3B Instruct Q4_K_M8K~3.8 GB · 69–108 tok/s estimateExcellent
Llama 3.2 3B Instruct IQ4_XS8K~3.7 GB · 75–117 tok/s estimateExcellent
Llama 3.2 3B Instruct Q3_K_M8K~3.5 GB · 79–125 tok/s estimateExcellent

Alternatives

GPUs that run Llama 3.2 3B Instruct Q4_K_M entirely in VRAM

Other models that fit the Arc B580

Where to buy a Arc B580

Prices and links follow the market you choose.

Intel Arc B580

12 GB GDDR6 · 456 GB/s · 190 W

ExcellentRuns Llama 3.2 3B Instruct Q4_K_M entirely in memory

Price not available

Availability unknown

View on amazon.com

Affiliate link

No current offer imported for this market: the link opens a marketplace search.

Frequently asked questions

Can the Arc B580 run Llama 3.2 3B Instruct?
The model needs about 3.8 GB and fits in the 11.6 GB available, with 7.8 GB to spare.
Which quantization should I use on the Arc B580?
Q8_0 is the highest-precision quantization that fits entirely in VRAM at 8K context. ~5.2 GB / 11.6 GB.
What is the largest context on the Arc B580?
At Q4_K_M, about 76K tokens fit entirely in VRAM with an f16 KV cache. A q8_0 KV cache roughly doubles that.
How fast is Llama 3.2 3B Instruct (Arc B580)?
We estimate 69–108 tok/s (estimate, confidence low). Based on: memory bandwidth (456 GB/s) and model size.