Skip to content
RfRigFit

What LLMs can the GMKtec EVO-X2 128GB (128 GB) run?

Mini PC

Data updated . Last reviewed the same day, when this catalog snapshot was written.

These Strix Halo boxes use soldered LPDDR5X at 256 GB/s. About 75% of the sticker memory can be assigned to the GPU, which is the allowance used here. Large models fit, then run slower than they would on a discrete GPU with more bandwidth.

Unified memory
128 GB
Usable
96.0 GB
Bandwidth
256 GB/s
List price
$3,499
Year
2025
Q4 models
42/49

GMKtec store price for 128GB / 1TB, 2026. About 96GB can be assigned to the GPU.

Check price on Amazon for GMKtec EVO-X2 128GBOpen in the checker

Amazon links are affiliate searches, not live prices.

Models

Sorted with Q4 fits first. Memory figures already include the 20% overhead.

Models GMKtec EVO-X2 128GB can run
ModelParamsQ4Q8FP16
Llama 3.2 1B Instruct
Llama
1.24BFits0.8 GBFits1.3 GBFits3.0 GB
Llama 3.2 3B Instruct
Llama
3.21BFits2.0 GBFits3.4 GBFits7.7 GB
Phi-3.5 mini
Phi
3.82BFits5.6 GBFits7.3 GBFits9.2 GB
Phi-4 mini
Phi
3.84BFits3.6 GBFits5.2 GBFits9.2 GB
Gemma 3 4B
Gemma
4.3BFits2.5 GBFits4.1 GBFits10.3 GB
Mistral 7B Instruct v0.3
Mistral
7.25BFits5.4 GBFits8.8 GBFits17.4 GB
Qwen2.5 7B Instruct
Qwen
7.62BFits5.2 GBFits8.6 GBFits18.3 GB
Gemma 4 E4B
Gemma
8BFits5.1 GBFits8.3 GBFits19.2 GB
DeepSeek R1 Distill Llama 8B
DeepSeek
8.03BFits6.0 GBFits9.6 GBFits19.3 GB
Llama 3.1 8B Instruct
Llama
8.03BFits4.9 GBFits9.6 GBFits19.3 GB
Qwen3 8B
Qwen
8.19BFits6.2 GBFits9.9 GBFits19.7 GB
Ministral 3 8B
Mistral
8.92BFits6.3 GBFits10.2 GBFits21.4 GB
Gemma 2 9B
Gemma
9.24BFits6.8 GBFits10.7 GBFits22.2 GB
Gemma 4 12B
Gemma
12BFits7.7 GBFits13.2 GBFits28.7 GB
Gemma 3 12B
Gemma
12.2BFits7.3 GBFits12.5 GBFits29.2 GB
Mistral Nemo 12B
Mistral
12.2BFits8.8 GBFits14.4 GBFits29.4 GB
Ministral 3 14B
Mistral
13.9BFits9.6 GBFits15.7 GBFits33.5 GB
Phi-4
Phi
14.7BFits10.6 GBFits17.3 GBFits35.2 GB
Qwen3 14B
Qwen
14.8BFits10.3 GBFits17.0 GBFits35.4 GB
Qwen2.5 14B Instruct
Qwen
14.8BFits10.6 GBFits17.3 GBFits35.4 GB
gpt-oss-20b
gpt-oss
20.9B
3.6B active
Fits11.8 GBFits12.3 GBFits50.2 GB
Mistral Small 3.2 24B
Mistral
24BFits15.7 GBFits26.4 GBFits57.6 GB
Gemma 4 26B-A4B
Gemma
25.8B
3.8B active
Fits17.3 GBFits27.2 GBFits61.9 GB
Gemma 2 27B
Gemma
27.2BFits16.6 GBFits30.0 GBFits65.3 GB
Gemma 3 27B
Gemma
27.4BFits16.5 GBFits28.7 GBFits65.8 GB
Qwen3.6 27B
Qwen
27.8BFits17.4 GBFits29.1 GBFits66.7 GB
Qwen3.8 27B
Qwen
27.8BFits17.0 GBFits29.6 GBFits66.7 GB
Qwen3 30B-A3B
Qwen
30.5B
3B active
Fits19.4 GBFits33.3 GBFits73.3 GB
GLM-4.7-Flash
GLM
31.2B
3B active
Fits18.8 GBFits32.3 GBFits74.9 GB
Gemma 4 31B
Gemma
31.3BFits19.7 GBFits34.0 GBFits75.1 GB
Qwen3 32B
Qwen
32.8BFits21.9 GBFits37.0 GBFits78.6 GB
DeepSeek R1 Distill Qwen 32B
DeepSeek
32.8BFits22.0 GBFits37.0 GBFits78.6 GB
Qwen2.5 32B Instruct
Qwen
32.8BFits22.0 GBFits37.0 GBFits78.6 GB
Qwen3.6 35B-A3B
Qwen
36B
3B active
Fits22.3 GBFits37.1 GBFits86.3 GB
Mixtral 8x7B
Mistral
46.7B
12.9B active
Fits27.5 GBFits50.7 GBNo112 GB
DeepSeek R1 Distill Llama 70B
DeepSeek
70.6BFits45.2 GBFits77.7 GBNo169 GB
Llama 3.3 70B Instruct
Llama
70.6BFits42.5 GBFits75.0 GBNo169 GB
Qwen2.5 72B Instruct
Qwen
72.7BFits46.7 GBFits80.2 GBNo174 GB
Llama 4 Scout
Llama
109B
17B active
Fits65.4 GBNo115 GBNo261 GB
gpt-oss-120b
gpt-oss
117B
5.1B active
Fits63.1 GBFits63.7 GBNo280 GB
Mistral Small 4
Mistral
119B
6.5B active
Fits74.0 GBNo127 GBNo287 GB
Mixtral 8x22B
Mistral
141B
39B active
Fits87.5 GBNo151 GBNo338 GB
Qwen3 235B-A22B
Qwen
235B
22B active
No144 GBNo252 GBNo564 GB
DeepSeek V4 Flash
DeepSeek
291B
13B active
No175 GBFits4.5 GBNo698 GB
Llama 4 Maverick
Llama
402B
17B active
No243 GBNo426 GBNo964 GB
Llama 3.1 405B Instruct
Llama
406BNo244 GBNo487 GBNo974 GB
Mistral Large 3
Mistral
675B
41B active
No407 GBNo716 GBNo1620 GB
DeepSeek R1
DeepSeek
684B
37B active
No405 GBNo714 GBNo1643 GB
DeepSeek V4 Pro
DeepSeek
1599B
49B active
No959 GBNo1673 GBNo3837 GB