Local AI machines, ranked by speed per dollar and euro
Complete machines only: Strix Halo mini PCs, DGX Spark and GB10 boxes, Mac Studios. Graphics cards are on a separate page. Prices checked 2026-10-16 14:35 UTC.
| # | Machine | USD | EUR | tok/s | tok/s per $1k | tok/s per €1k |
|---|---|---|---|---|---|---|
| 1 | Mac Studio M5 Max 40-core GPU, 128GB / 512GB Apple · 128GB |
$5,099 | €5,909 | 79 | 15.5 | 13.4 |
| 2 | Framework Desktop 128GB (DIY) Strix Halo · 128GB |
$3,449 | €3,889 | 53 | 15.3 | 13.6 |
| 3 | GMKtec EVO-X2 128GB / 1TB Strix Halo · 128GB |
$3,500 | €3,300 | 53 | 15.1 | 16.0 |
| 4 | GMKtec EVO-X2 128GB / 2TB Strix Halo · 128GB |
$3,650 | €3,450 | 53 | 14.4 | 15.3 |
| 5 | GMKtec EVO-X3 128GB / 2TB Strix Halo · 128GB |
$3,800 | €3,550 | 53 | 13.9 | 14.9 |
| 6 | Minisforum MS-S1 MAX 128GB / 2TB Strix Halo · 128GB |
$3,879 | ≈€3,447 | 53 | 13.6 | 15.3 |
| 7 | NVIDIA DGX Spark (Founders Edition) DGX Spark / GB10 · 128GB |
$4,699 | ≈€4,176 | 59 | 12.5 | 14.1 |
| 8 | Beelink GTR9 Pro 128GB / 2TB Strix Halo · 128GB |
$4,349 | ≈€3,865 | 53 | 12.1 | 13.6 |
| 9 | Mac Studio M5 Ultra 80-core GPU, 256GB / 1TB Apple · 256GB |
$10,799 | €12,529 | 130 | 12.0 | 10.4 |
| 10 | MacBook Pro 16" M5 Max 40-core GPU, 128GB / 2TB Apple · 128GB |
$6,999 | €7,999 | 79 | 11.3 | 9.9 |
| 11 | Dell Pro Max / Pro Precision with GB10, 128GB / 2TB DGX Spark / GB10 · 128GB |
$5,903 | ≈€5,246 | 59 | 9.9 | 11.2 |
| 12 | ASUS Ascent GX10 (GB10) DGX Spark / GB10 · 128GB |
$5,999 | ≈€5,331 | 59 | 9.8 | 11.0 |
| GMKtec EVO-X2 64GB / 1TB Strix Halo · 64GB |
$2,200 | €2,000 | doesn't fit in memory | – | – | |
| Minisforum MS-S1 MAX 64GB / 2TB Strix Halo · 64GB |
$2,599 | ≈€2,310 | doesn't fit in memory | – | – | |
| Framework Desktop 64GB (DIY) Strix Halo · 64GB |
≈$2,486 | €2,209 | doesn't fit in memory | – | – | |
| HP Z2 Mini G1a (Ryzen AI Max+ PRO 395) Strix Halo · 128GB |
n/a | n/a | 53 | – | – | |
| MSI EdgeXpert MS-C931 (GB10) DGX Spark / GB10 · 128GB |
n/a | n/a | 59 | – | – | |
| Lenovo ThinkStation PGX (GB10) DGX Spark / GB10 · 128GB |
n/a | n/a | 59 | – | – | |
| Mac Studio M5 Ultra 64-core GPU, 96GB / 1TB Apple · 96GB |
$5,499 | €6,699 | no public result yet | – | – | |
| Mac mini M5 Pro 16-core GPU, 64GB / 512GB Apple · 64GB |
$2,699 | €3,129 | doesn't fit in memory | – | – | |
| Mac mini M5 Pro 20-core GPU, 64GB / 512GB Apple · 64GB |
$2,899 | €3,349 | doesn't fit in memory | – | – |
Large model: gpt-oss-120b MXFP4 (59 GiB weights), text generation tok/s at empty context, single request. Needs about 59 GiB for weights plus context and OS, so devices with 64GB or less are marked 'doesn't fit'. Sources and engines differ per chip (listed on Benchmarks); Apple figures are from Ollama/LM Studio, not llama-bench.
≈ = converted at the ECB rate.
Rather rent?
See what an hour on a cloud GPU costs, and how many hours equal one box.
Rent-instead comparison →Rank by another model
Qwen3.8 27B · Qwen3.6 35B-A3B · Gemma 4 12B · Qwen3-Coder 30B-A3B · Qwen3.5 9B · Qwen3.6 27B · Gemma 4 26B-A4B · Muse Glimmer 30B · gpt-oss-20b · gpt-oss-120b · Qwen3 8B · Llama 3.1 8B · Qwen3 32B · Llama 3.3 70B · Qwen3.8-Flash-Next · GLM-5.3-Flash