GMKtec EVO-X2 Ryzen AI Max+ 395 128GB vs Mac Studio M4 Max 128GB
For running models locally, priced in Canada. At 8k context the first runs 41 models fully in memory, the second 41.
Specifications
| Spec | GMKtec EVO-X2 Ryzen AI Max+ 395 128GB | Mac Studio M4 Max 128GB |
|---|---|---|
| Chip | Ryzen AI Max+ 395 (Radeon 8060S) | Apple M4 Max (40-core GPU) |
| Memory bus | 256-bit | 512-bit |
| Memory bandwidth | 256 GB/s | 546 GB/s |
| FP16 compute | 59.4 TFLOPS | 16.16 TFLOPS |
| Software | ROCm | Metal |
| Unified memory | 128 GB | 128 GB |
| GPU can use by default | 96 GB | 96 GB |
| GPU can use, tuned | 120 GB | 120 GB |
| Operating system | windows | macos |
| Processor | Ryzen AI Max+ 395 | M4 Max 16-core |
| Storage | 2000 GB | 1000 GB |
What each runs
Estimated decode speed with the most precise quant that fits each one est.
| Model | GMKtec EVO-X2 Ryzen AI Max+ 395 128GB | Mac Studio M4 Max 128GB |
|---|---|---|
| DeepSeek V4 Flash | 11tok/sUD-Q2_K_XL | 28tok/sUD-Q2_K_XL |
| Qwen 3.8 Flash Next | 3tok/sUD-Q3_K_XL | 7tok/sUD-Q3_K_XL |
| MiniMax M2.7 | 15tok/sUD-Q2_K_XL | 37tok/sUD-Q2_K_XL |
| Qwen 3.8 27B | 5tok/sQ8_0 | 12tok/sQ8_0 |
| Qwen 3.5 122B-A10B | 10tok/sQ6_K | 25tok/sQ6_K |
| Qwen 3.6 27B | 5tok/sQ8_0 | 12tok/sQ8_0 |
| gpt-oss 120B | 25tok/sMXFP4 | 62tok/sMXFP4 |
| Qwen3 Coder Next | 19tok/sQ8_0 | 48tok/sQ8_0 |
| Nemotron 3 Super 120B-A12B | 10tok/sUD-Q4_K_M | 26tok/sUD-Q4_K_M |
| Gemma 4 31B | 4tok/sQ8_0 | 10tok/sQ8_0 |
| Mistral Small 4 119B | 14tok/sUD-Q6_K | 35tok/sUD-Q6_K |
| Qwen 3.6 35B-A3B | 26tok/sQ8_0 | 64tok/sQ8_0 |
| Qwen 3.5 27B | 5tok/sQ8_0 | 12tok/sQ8_0 |
| Gemma 4 26B-A4B | 19tok/sQ8_0 | 47tok/sQ8_0 |
| GLM 4.5 Air | 5tok/sQ6_K | 13tok/sQ6_K |
| Qwen 3.5 35B-A3B | 26tok/sQ8_0 | 64tok/sQ8_0 |
| Qwen3 Next 80B-A3B | 25tok/sQ8_0 | 63tok/sQ8_0 |
| GLM 4.7 Flash | 15tok/sQ8_0 | 37tok/sQ8_0 |
| Devstral Small 2 24B | 5tok/sQ8_0 | 13tok/sQ8_0 |
| Nemotron 3.5 Lightning 30B-A3B | 25tok/sUD-Q6_K_XL | 62tok/sUD-Q6_K_XL |
| gpt-oss 20B | 32tok/sMXFP4 | 81tok/sMXFP4 |
| Kimi Linear 48B-A3B | 26tok/sQ8_0 | 65tok/sQ8_0 |
| Qwen3 Coder 30B-A3B | 21tok/sQ8_0 | 53tok/sQ8_0 |
| Qwen3 32B | 4tok/sQ8_0 | 10tok/sQ8_0 |
| Gemma 4 12B | 11tok/sQ8_0 | 27tok/sQ8_0 |
| Qwen 3.5 9B | 14tok/sQ8_0 | 36tok/sQ8_0 |
| Mistral Small 3.2 24B | 5tok/sQ8_0 | 13tok/sQ8_0 |
| Gemma 3 27B | 5tok/sQ8_0 | 12tok/sQ8_0 |
| DeepSeek R1 Distill Qwen 32B | 4tok/sQ8_0 | 10tok/sQ8_0 |
| Llama 3.3 70B | 2tok/sQ8_0 | 5tok/sQ8_0 |
| Ministral 3 14B | 9tok/sQ8_0 | 23tok/sQ8_0 |
| Qwen3 14B | 8tok/sQ8_0 | 21tok/sQ8_0 |
| DeepSeek R1 0528 Qwen3 8B | 14tok/sQ8_0 | 36tok/sQ8_0 |
| Qwen3 8B | 14tok/sQ8_0 | 36tok/sQ8_0 |
| Phi-4 14B | 8tok/sQ8_0 | 21tok/sQ8_0 |
| Qwen 3.5 4B | 29tok/sQ8_0 | 74tok/sQ8_0 |
| Ministral 3 8B | 14tok/sQ8_0 | 35tok/sQ8_0 |
| Granite 4.2 8B | 13tok/sQ8_0 | 33tok/sQ8_0 |
| Gemma 4 E4B | 17tok/sQ8_0 | 42tok/sQ8_0 |
| LFM2.5 8B-A1B | 58tok/sQ8_0 | 145tok/sQ8_0 |
| Llama 3.1 8B | 18tok/sQ6_K | 46tok/sQ6_K |
| GLM 5.3 Flash | won't run | won't run |