Best Macs and mini PCs for local AI
Ranked by what each one runs fully in memory at 32k context, with prices in the US today.
In short
The top pick in the US right now is the Mac Studio M3 Ultra 256GB, $8,499 at eBay: it runs DeepSeek V4 Flash fully in memory at 32k context at about 23 tok/s (estimated). Ranked by the most capable open model each one runs fully in memory at 32k context and a usable speed, then by how many models it runs, then by speed on Qwen 3.8 27B, then by price.
Updated . Speeds are estimates; how they are worked out.
The list
Best first. Speeds are estimates est.
| # | Machine | Memory | Price | Best model it runs at 32k | Models in memory | Qwen 3.8 27B | tok/s per $1,000 |
|---|---|---|---|---|---|---|---|
| 1 | Mac Studio M3 Ultra 256GB | 256 GB | $8,499eBay | DeepSeek V4 FlashUD-Q4_K_XL23 tok/s | 36 | 17 tok/s | 2 |
| 2 | Mac Studio M3 Ultra 512GB | 512 GB | $23,000eBay | DeepSeek V4 FlashUD-Q4_K_XL23 tok/s | 35 | 17 tok/s | 0.7 |
| 3 | Mac Studio M4 Max 128GB | 128 GB | $4,999eBay | DeepSeek V4 FlashUD-Q2_K_XL22 tok/s | 29 | 11 tok/s | 2.3 |
| 4 | Mac Studio M3 Ultra 96GB | 96 GB | $4,399eBay | Qwen 3.8 27BQ8_017 tok/s | 34 | 17 tok/s | 3.9 |
| 5 | Mac Studio M4 Max 64GB | 64 GB | $3,400eBay | Qwen 3.5 122B-A10BUD-Q2_K_XL50 tok/s | 25 | 11 tok/s | 3.3 |
| 6 | Mac mini M4 Pro 64GB | 64 GB | $2,900eBay | Qwen 3.5 122B-A10BUD-Q2_K_XL25 tok/s | 17 | 6 tok/s | 2 |
| 7 | ASUS Ascent GX10 128GB | 128 GB | $6,100eBay | gpt-oss 120BMXFP423 tok/s | 16 | 5 tok/s | 0.9 |
| 8 | NVIDIA DGX Spark 128GB | 128 GB | $7,399eBay | gpt-oss 120BMXFP423 tok/s | 16 | 5 tok/s | 0.7 |
| 9 | GMKtec EVO-X2 Ryzen AI Max+ 395 128GB | 128 GB | $3,650eBay | gpt-oss 120BMXFP420 tok/s | 12 | 5 tok/s | 1.2 |
| 10 | Mac mini M4 Pro 48GB | 48 GB | $2,149eBay | Qwen 3.6 35B-A3BQ8_028 tok/s | 13 | 6 tok/s | 2.7 |
| 11 | Mac mini M4 32GB | 32 GB | $1,420eBay | LFM2.5 8B-A1BQ8_018 tok/s | 1 | 4 tok/s | 3 |
| 12 | Mac mini M4 16GB | 16 GB | $749eBay | LFM2.5 8B-A1BQ8_018 tok/s | 1 | won't run | – |
Questions
- What is the best Mac or mini PC for running AI models locally?
- The top pick in the US right now is the Mac Studio M3 Ultra 256GB, $8,499 at eBay: it runs DeepSeek V4 Flash fully in memory at 32k context at about 23 tok/s (estimated).
- Which gives the most speed for the money?
- The Mac Studio M3 Ultra 96GB: about 17 tok/s on Qwen 3.8 27B for $4,399, the most tokens per second per unit of price on this list (estimated).
- What is the cheapest option here that runs Qwen 3.8 27B fully in memory?
- The Mac mini M4 32GB, $1,420 at eBay, at about 4 tok/s at 32k context (estimated).