GMKtec EVO-X5 Pro AMD Ryzen™ AI Max+ PRO 495 AI Mini PC - GMKtec Skip to content

⚡Price increase coming soon — order now to lock in the current price!

3.jpg__PID:df02dc92-fec6-47c3-a1d9-1249d865872b
AMD Ryzen AI Max+ PRO 495 

GMKtec EVO-X5 Pro 

Next-Ceneration Desktop Al Supercomputing Workstation
The World's First Agentic PC capable of Running 320B-parameter LLMs Fully offline

GMKtec
GMKtec
GMKtec
GMKtec

AI Model Performance

Model / Quantization Model Size GB Decode tok/s Prefill tok/s TTFT s
Qwen3.6-35B-A3B ROCmFP4-STRIX · Chat 19.86 79.20 1240.91 0.462
Qwen3.6-35B-A3B UD-Q4_K_XL 22.85 81.77 1063.82 0.935
Qwen3.8-Flash-Next ROCmFP4-FAST 93.48 45.08 391.81 2.980
Qwen3.8-Flash-Next UD-Q4_K_XL 111.33 23.49 283.56 2.033
Qwen3.8-27B Q4_K_M 17.10 21.01 252.39 4.457
GLM-5.3-Flash UD-IQ4_XS 146.05 17.79 70.18 14.71
DeepSeek-V4-Flash-0731 UD-Q4_K_XL 144.44 17.66 104.86 9.770

Decode: token generation speed. Prefill: prompt processing speed. TTFT: time to first token.
Source: GMKtec Laboratory. Actual performance varies with model, quantization, context length and software configuration.

GMKtec
GMKtec
GMKtec
GMKtec

MOC Fingerprint Security

Fast and secure biometric authentication prevents unauthorized access 

and protects sensitive data, critical files and intellectual property across enterprise, 

financial, healthcare and AI workstation environments.

GMKtec
GMKtec
GMKtec

Keep Your AI Local.
Keep Running Costs Down.

Explore how local AI can reduce recurring costs. The examples below illustrate two separate scenarios: replacing a workstation and replacing cloud API usage.

Scenario 01 · Electricity

Lower Power. Lower Running Costs.

$42.03 / month

Potential electricity savings in an illustrative 180W versus 600W comparison. These are assumed whole-system power values, not measured EVO-X5 Pro results.

720 operating hours per month · $0.139/kWh · USD
Example System Energy / Month Cost / Month
180W system 129.6 kWh $18.01
600W workstation 432.0 kWh $60.05
Potential savings* 302.4 kWh $42.03

*Calculated before rounding. A valid product comparison requires measured energy use for equivalent completed workloads. Equal operating hours do not establish equal inference output.

Scenario 02 · Cloud API

Local Inference. Fewer API Charges.

$360 / month

Potential cloud API charges avoided if the example usage below is fully replaced by suitable local inference. Rates are illustrative.

Example monthly cloud usage and pricing · USD
Token Type Monthly Usage Rate / 1M Cost
Input 100M $2 $200
Output 20M $8 $160
Potential API charges avoided $360

Producing 20M output tokens in 30 days requires approximately 7.72 output tokens per second averaged continuously across the month. Actual capacity must be verified for the chosen model, context length and workload.

Illustrative calculations only, not measured EVO-X5 Pro savings or performance claims. The two scenarios use different baselines and should not be added together. Actual savings depend on measured wall power, workload, model quality, throughput, utilization, electricity rates and API pricing. Adapter ratings and processor power settings are not whole-system power measurements. API charges avoided are not net savings: hardware, local electricity, software and maintenance costs must also be considered.

GMKtec
GMKtec
GMKtec

I/O

GMKtec
  • 5 stars: 4 (80%)
  • 4 stars: 0 (0%)
  • 3 stars: 0 (0%)
  • 2 stars: 0 (0%)
  • 1 star: 1 (20%)
A
Anonymous

Solid build. Heavier than I thought. Love the support extended through every step on discord!

E
Evgeny Matveev

A grat one. I like it. Very quiet. Runs qwen 3.8 flash next q8_0 at 15 tokens in ubuntu 26.04 via lm studio. I have some iperfect finish on the case, but I dont really care as it is work LM machine.

A
Anonymous

Great Product So Far. Only small complaint is that you guys had to use those star screws, which was harder to find

A
Anonymous

Great product and fast shipping!

A
Anonymous

Q