Can Llama 3.2 1B Instruct run on the Arc B580?
Yes. At 8K context the best fit is F16, using about 5.5 GB of the 11.4 GB usable on the Arc B580. Estimated speed ceiling: ~92 tokens/sec.
Memory by quant and context
| Quant | 4K ctx | 8K ctx | 32K ctx | 128K ctx | Max tok/s (est., 8K) |
|---|---|---|---|---|---|
| 2-bit (Q2_K) • | 1.3 GB Fits | 1.4 GB Fits | 2.2 GB Fits | 5.2 GB Fits | 785 |
| 3-bit (Q3_K_M) • | 1.4 GB Fits | 1.5 GB Fits | 2.3 GB Fits | 5.3 GB Fits | 660 |
| 4-bit (Q4_K_M) • | 1.5 GB Fits | 1.6 GB Fits | 2.4 GB Fits | 5.4 GB Fits | 565 |
| 5-bit (Q5_K_M) • | 1.6 GB Fits | 1.7 GB Fits | 2.5 GB Fits | 5.5 GB Fits | 500 |
| 6-bit (Q6_K) • | 1.7 GB Fits | 1.8 GB Fits | 2.6 GB Fits | 5.6 GB Fits | 446 |
| 8-bit (Q8_0) • | 2.0 GB Fits | 2.1 GB Fits | 2.9 GB Fits | 5.9 GB Fits | 345 |
| 16-bit (F16/BF16) • | 5.4 GB Fits | 5.5 GB Fits | 6.3 GB Fits | 9.3 GB Fits | 92 |
• real GGUF file size from unsloth/Llama-3.2-1B-Instruct-GGUF. Speed is a bandwidth ceiling (456 GB/s), not a benchmark. Methodology.
More
Llama 3.2 1B Instruct on every GPU and Mac · Every model that runs on the Arc B580 · Other models that fit 12 GB VRAM · Try other settings in the calculator
Data: Hugging Face Hub (meta-llama/Llama-3.2-1B-Instruct), Intel specs. Data as of 2026-11-17.