Best mini PC for local LLMs
Our pick for a mini PC that runs big local models is the Framework Desktop 128GB (DIY): $3,449 / €3,889, 52.7 tok/s on gpt-oss-120b, which works out to 15.3 tok/s per $1,000. Runner-up: GMKtec EVO-X2 128GB / 1TB at $3,500 / €3,300 and 52.7 tok/s.
Mini PCs here means Strix Halo and DGX Spark/GB10 boxes with 128GB, ranked on gpt-oss-120b, a model that needs about 64GB.
| # | Machine | Price | tok/s | per $1k |
|---|---|---|---|---|
| 1 | Framework Desktop 128GB (DIY) Strix Halo · 128GB · checked 2026-10-16 | $3,449 €3,889 | 52.7 | 15.3 |
| 2 | GMKtec EVO-X2 128GB / 1TB Strix Halo · 128GB · checked 2026-10-16 | $3,500 €3,300 | 52.7 | 15.1 |
| 3 | GMKtec EVO-X2 128GB / 2TB Strix Halo · 128GB · checked 2026-10-16 | $3,650 €3,450 | 52.7 | 14.4 |
| 4 | GMKtec EVO-X3 128GB / 2TB Strix Halo · 128GB · checked 2026-10-16 | $3,800 €3,550 | 52.7 | 13.9 |
| 5 | Minisforum MS-S1 MAX 128GB / 2TB Strix Halo · 128GB · checked 2026-10-16 | $3,879 ≈€3,447 | 52.7 | 13.6 |
| 6 | NVIDIA DGX Spark DGX Spark / GB10 · 128GB · checked 2026-10-16 | $4,699 ≈€4,176 | 58.7 | 12.5 |
| 7 | Beelink GTR9 Pro 128GB / 2TB Strix Halo · 128GB · checked 2026-10-16 | $4,349 ≈€3,865 | 52.7 | 12.1 |
| 8 | Dell Pro Max / Pro Precision with GB10 DGX Spark / GB10 · 128GB · checked 2026-10-16 | $5,903 ≈€5,246 | 58.7 | 9.9 |
tok/s = generation speed on gpt-oss-120b from public benchmarks. per $1k = tok/s per $1,000 of the US price. ≈ = converted at the ECB rate.
Not ranked: GMKtec EVO-X2 64GB / 1TB (doesn't fit in memory); Minisforum MS-S1 MAX 64GB / 2TB (doesn't fit in memory); Framework Desktop 64GB (DIY) (doesn't fit in memory); HP Z2 Mini G1a (Ryzen AI Max+ PRO 395) (no tracked price); MSI EdgeXpert MS-C931 (GB10) (no tracked price); Lenovo ThinkStation PGX (GB10) (no tracked price).
How we picked
Ranked by measured tokens per second per $1,000 of the current US price. Speeds are public benchmark results, one per chip, linked on each product page and on Benchmarks; engines and quants differ between sources. Prices checked 2026-10-16 19:02 UTC and refreshed every 6 hours.
FAQ
Which mini PC is best for running LLMs locally?
On current prices and public gpt-oss-120b results, Framework Desktop 128GB (DIY). It costs $3,449 / €3,889 and does 52.7 tok/s.
Is DGX Spark faster than Strix Halo?
On gpt-oss-120b, the public results we use show DGX Spark / GB10 58.7 tok/s, Strix Halo 52.7 tok/s. Engines differ between sources.
Do 64GB mini PCs run gpt-oss-120b?
No. gpt-oss-120b needs about 64GB for weights alone, so we only count 128GB machines.