Mac mini M5 Pro, 24GB vs Strix Halo GMKtec EVO-X2, 128GB for local AI
The GMKtec EVO-X2, 128GB holds 39 of the 48 current models here, and the Mac mini M5 Pro, 24GB holds 16. The Mac mini M5 Pro, 24GB costs $1,801 less. On Gemma 4 12B, the strongest model both hold, the Mac mini M5 Pro, 24GB is about 1.2× faster: 29 tok/s against 24, both estimated from memory bandwidth. The Mac mini M5 Pro, 24GB pays for itself sooner, in 135 years against 267 years at 500k tokens a day, though that is each machine on its own strongest model rather than on the same one.
| Mac mini M5 Pro, 24GB | Strix Halo GMKtec EVO-X2, 128GB | |
|---|---|---|
| Price | $1,699 | $3,500 |
| Memory | 24 GB | 128 GB |
| Usable by the GPU | 16 GB | 96 GB |
| Memory bandwidth | 307 GB/s | 256 GB/s |
| Power under load | 140 Wstand-in | 150 W |
| Models that fit | 16 | 39 |
| Best model it runs | Gemma 4 12B | Qwen3.8 27B |
| Speed on that model | 29 tok/s estimatedGemma 4 12B | 10 tok/s estimatedQwen3.8 27B |
| Pay-back on that model | Pays back in 135 years | Pays back in 267 years |
The 140 W beside the Mac mini M5 Pro, 24GB is a stand-in, not a figure for that machine: the data borrows it from the nearest hardware it does have, and the machine's own page names which and why. So the two figures above are not like for like, and the electricity in its pay-back here is priced from a borrowed number.
Run the numbers on the Mac mini M5 Pro, 24GB · or the Strix Halo GMKtec EVO-X2, 128GB
Jump to: Side by side on Gemma 4 12B · How much use it takes to pay back · What the extra memory buys · The assumptions behind both columns
Side by side on Gemma 4 12B
The table above gives each machine the strongest model it can hold, and those are not the same model, so the two speeds in it are not a race. Gemma 4 12B is the strongest model both machines hold, so this is the pair running the same work.
| Mac mini M5 Pro, 24GB | Strix Halo GMKtec EVO-X2, 128GB | |
|---|---|---|
| Speed | 29 tok/s estimated | 24 tok/s estimated |
| Pay-back | Pays back in 135 years | Pays back in 296 years |
Run Gemma 4 12B on the Mac mini M5 Pro, 24GB · or on the Strix Halo GMKtec EVO-X2, 128GB
How much use it takes to pay back
Everything above is at 500k tokens a day. Pay-back moves with how much you actually run, so here are both machines on the same model, at the five levels of use the calculator names. The Mac mini M5 Pro, 24GB pays back sooner at every level of use, so this is not a choice that turns on how hard you work it.
| A day's use | Mac mini M5 Pro, 24GB | Strix Halo GMKtec EVO-X2, 128GB |
|---|---|---|
| 50ka few chats a day | 1,353 years | 2,962 years |
| 200klight assistant use | 338 years | 741 years |
| 1Ma moderate coding-assistant day | 68 years | 148 years |
| 4Mheavy coding with an agent | 17 years | 37 years |
| 20Magents running most of the day | 3.4 years | 7.4 years |
Run the Mac mini M5 Pro, 24GB at 20M tokens a day · or the Strix Halo GMKtec EVO-X2, 128GB
What the extra memory buys
The Strix Halo GMKtec EVO-X2, 128GB holds 23 models the Mac mini M5 Pro, 24GB cannot at 32k of context. The strongest of them are what the difference in memory actually buys.
| Model | Weights | Needs at 32k | On the Strix Halo GMKtec EVO-X2, 128GB |
|---|---|---|---|
| Qwen3.8 27BSonnet-class | 16 GB | 19 GB | 10 tok/s estimated |
| Ling 3.0 flashHaiku-class | 78 GB | 78 GB | 41 tok/s estimated |
| Qwen3.6 27BHaiku-class | 17 GB | 19 GB | 10 tok/s estimated |
| Nemotron 3.5 Bolt 74B-A9BHaiku-class | 22 GB | 22 GB | 48 tok/s estimated |
| Qwen3.6 35B-A3BHaiku-class | 22 GB | 23 GB | 56 tok/s estimated |
| Muse Glimmer 30BHaiku-class | 17 GB | 18 GB | 11 tok/s estimated |
17 more, on the Strix Halo GMKtec EVO-X2, 128GB page.
The extra memory buys context as well. Of the 16 current models both machines hold at 32k, 4 run to a longer window on the Strix Halo GMKtec EVO-X2, 128GB: the weights are a fixed size and the KV cache is not, so what the weights leave spare is what a longer context grows into. Ministral 3 14B reaches 256k there against 32k on the Mac mini M5 Pro, 24GB, each the longest window the calculator offers that the machine still holds it at.
The Mac mini M5 Pro, 24GB is also head to head with another computer: Mac Studio M5 Max, 128GB · DGX Spark, 128GB · MacBook Air M5 (15-inch), 16GB · MacBook Pro M5 Pro (16-inch), 64GB · Corsair AI Workstation 300, 64GB. With a graphics card: RTX PRO 6000 Blackwell, 96GB · Radeon AI PRO R9700, 32GB. With the same machine at another memory size: 48GB · 64GB. With the machine it replaced: Mac mini M4 Pro, 24GB.
The GMKtec EVO-X2, 128GB is also head to head with another computer: Mac Studio M5 Max, 128GB · DGX Spark, 128GB · MacBook Air M5 (15-inch), 16GB · MacBook Pro M5 Pro (16-inch), 64GB · Framework Desktop, 128GB · Mac Studio M5 Max, 64GB. With a graphics card: RTX PRO 6000 Blackwell, 96GB · Radeon AI PRO R9700, 32GB. With the same machine at another memory size: 64GB.
The assumptions behind both columns
Both columns use the same usage: 500k tokens a day at 15:1 input to output, 32k of context, $0.17 per kWh, and today's API prices held flat. Speeds marked estimated are worked out from memory bandwidth rather than measured, and pay-back scales with them. Change any of it in the calculator.
More head to head: everything the Mac mini M5 Pro, 24GB runs · everything the Strix Halo GMKtec EVO-X2, 128GB runs · every other match-up · the quickest pay-back at each level of use · every model against the frontier