Question 1

Can the Radeon Vega 3 (iGPU) run DEEPSEEK-R1 distill 32B?

Accepted Answer

DEEPSEEK-R1 distill 32B needs about 21GB working set (weights 20GB + KV cache 1GB at 4k context). The Radeon Vega 3 (iGPU) offers ~5.5GB usable — verdict: no. 21.0 GB vs 5.5 GB usable — does not fit

Question 2

How fast is DEEPSEEK-R1 distill 32B on the Radeon Vega 3 (iGPU)?

Accepted Answer

~0.48–1.5 tok/s on the Radeon Vega 3 (iGPU). computed roofline: 38.4 GB/s × 0.25–0.80 efficiency window / 20.0GB (Q4_K_M) — bands, never points.

Question 3

What should I run DEEPSEEK-R1 distill 32B on instead?

Accepted Answer

Mac mini M4 (32GB) runs it fully ($999).

Can the Radeon Vega 3 (iGPU) run DEEPSEEK-R1 distill 32B?

Run DEEPSEEK-R1 distill 32B on instead