GMKtec EVO-T2 Mini PC Packs 180 TOPS for Local AI

GMKtec has introduced the EVO-T2, a compact mini PC aimed at handling demanding AI workloads on the desktop or at the edge. The device purports up to 180 TOPS of performance, positioning itself as a bridge between ultraportable small form factors and more power‑hungry AI workstations.

Under the hood the EVO-T2 supports fast PCIe storage and a broad selection of connectivity options, including 10GbE and USB4, which should appeal to prosumers and developers who move large datasets around or need low‑latency networking for inference tasks. There’s also support for OCuLink, giving the system expandability that many other mini PCs lack.

One of the more interesting features is an SSD‑assisted memory mode — a pseudo‑memory system that lets the EVO‑T2 offload portions of model state to high‑speed NVMe drives. That approach doesn’t replace DRAM or dedicated accelerator memory, but it can reduce the footprint required in RAM and make larger models more practical on a compact platform. It’s a pragmatic trade‑off for users wanting local model execution without investing in full‑scale servers.

Performance claims such as 180 TOPS are meaningful but should be evaluated in the context of real‑world workloads, model architectures and software optimizations. As with many specialized systems, drivers, runtime support and software integration will determine how smoothly the EVO‑T2 slots into existing AI development pipelines.

For anyone experimenting with local inference, real‑time analytics or edge AI deployment, the EVO‑T2 is an intriguing option. It blends dense compute figures with practical I/O and storage features, offering a route to run larger models closer to the data without a traditional rack setup. Keep an eye out for hands‑on tests and benchmarks to see how it performs outside of marketing numbers.