Best Macs and mini PCs for local AI

Ranked by what each one runs fully in memory at 32k context, with prices in the UK today.

In short

The top pick in the UK right now is the Mac Studio M3 Ultra 96GB, £5,499 at eBay: it runs Qwen 3.8 27B fully in memory at 32k context at about 17 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.

#MachineMemoryPriceBest model it runs at 32kModels in memoryQwen 3.8 27Btok/s per £1,000
1Mac Studio M3 Ultra 96GB96 GB£5,499eBayQwen 3.8 27BQ8_017 tok/s3417 tok/s3.1
2Mac Studio M4 Max 64GB64 GB£3,399eBayQwen 3.5 122B-A10BUD-Q2_K_XL50 tok/s2511 tok/s3.3
3GMKtec EVO-X2 Ryzen AI Max+ 395 128GB128 GB£3,320eBaygpt-oss 120BMXFP420 tok/s125 tok/s1.4
4Mac mini M4 Pro 48GB48 GB£3,000eBayQwen 3.6 35B-A3BQ8_028 tok/s136 tok/s1.9
5Mac mini M4 16GB16 GB£765eBayLFM2.5 8B-A1BQ8_018 tok/s1won't run–

Questions

What is the best Mac or mini PC for running AI models locally?
The top pick in the UK right now is the Mac Studio M3 Ultra 96GB, £5,499 at eBay: it runs Qwen 3.8 27B fully in memory at 32k context at about 17 tok/s (estimated).
Which gives the most speed for the money?
The Mac Studio M4 Max 64GB: about 11 tok/s on Qwen 3.8 27B for £3,399, 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 Pro 48GB, £3,000 at eBay, at about 6 tok/s at 32k context (estimated).

Other lists