Local AI Stats.
Skip to content

COMPUTER RECORD · SOURCED SPECIFICATIONS

GMKtec EVO-T2S · 64GB: specifications and local LLMs

64 GB system RAM. 64GB onboard memory; Arc B390 graphics; 10GbE and OCuLink.

Catalog reviewed . Listing evidence: 2026-09-30. Price and performance records retain their own dates.

As an Amazon Associate I earn from qualifying purchases.

CPU, GPU, memory and storage

CPU
Core Ultra X7 358H · 16 cores / 16 threads
GPU
Intel Arc B390 · 12 Xe cores
Installed memory
64 GB · LPDDR5X-8533 · Amazon.com ASIN listing
GPU memory pool
Shared allocation not verified
Memory bandwidth
Not verified for this installed configuration
Storage
1000 GB · Amazon.com ASIN listing
Operating system
Windows 11 Pro (manufacturer; seller license unverified)
Expansion
Dual M.2; OCuLink PCIe 4.0 x4; dual USB4; 10GbE + 2.5GbE
NPU advertised peak
50 TOPS · not LLM tokens/s
Whole-chip advertised peak
180 TOPS · distinct from NPU-only performance

Non-upgradeable RAM; no verified local LLM throughput for this listing.

Amazon advertises a PCIe 5.0 1TB SSD, while GMKtec specifies PCIe 4.0 for its 1TB option. Installed SSD interface is unresolved. This is not the 853GB aiDAPTIV+ SSD configuration.

Selected ASIN and configuration matched the indexed Amazon.com product title on 30 September 2026. The retrieved page omitted the main offer price; this is not a live price or stock check. Current seller, stock and checkout price can change.

What local models could GMKtec EVO-T2S · 64GB run?

The 64 GB system-RAM pool gives a 44.8 GB initial weight budget using our 30% (minimum 2 GB) reserve. This is a CPU-inference budget. Integrated-GPU allocation is not established.

127 catalog files pass this size check. The largest is Midnight-Miqu-70B-v1.5 at 41.423 GB (Q4_K_M). This is a catalog weight limit, not a measured maximum practical model size. Context length, runtime and architecture support can rule out a candidate.

Model snapshot: 14 October 2026. The examples below use MMLU-Pro; models without that score are not ranked in this table.

MMLU-Pro memory candidates · GB = 1 billion bytes. Published scores may use different settings or precision.
Model / quantizationWeightsParametersScore / 100Score / GBBudget leftSources
Qwen3.5-4BQ4_K_M · Pareto choice2.741 GB4.206B79.128.85942.059 GBGet weights
Score source
MiniCPM5 2BQ4_K_M · Pareto choice1.561 GB2.517B70.845.34643.239 GBGet weights
Score source
Qwen3.5-9BQ4_K_M · Pareto choice5.681 GB9.653B82.514.52339.119 GBGet weights
Score source
LFM2.5 230MQ4_K_M · Pareto choice0.153 GB0.23B20.25132.00244.647 GBGet weights
Score source
Qwen3.5-27BQ4_K_M · Pareto choice16.741 GB27.781B86.15.14328.059 GBGet weights
Score source

Choose this computer in the memory tool · Compare its memory tier

Published LLM performance

Local AI Stats has not tested this computer. No matching token-speed result is recorded for this configuration.

Advertised compute peaks and bandwidth are not tokens per second. Compare the same model, quantization, context and concurrency.

Price and value

ASIN: B0H1G9Y1FZ · Amazon.com listing

$2,399.99 USD reference. GMKtec US reference observed 30 Sep 2026: X7 358H, 64GB RAM, 1TB SSD, US plug, USD 2,399.99. Not an Amazon offer. The Amazon SSD-generation claim differs from the manufacturer specification.

Reference price per installed-memory GB: $37.5. This measures cost per GB, not speed or total running cost.

As an Amazon Associate I earn from qualifying purchases. Check the exact configuration, seller, marketplace and current price before buying.

GET NOW · Amazon

Related hardware

Compare all hardware · Compare budgets

Specification and price sources