GMKtec EVO T2 with 180 TOPS, Panther Lake as the Technical Foundation of the AI Mini PC

Key Takeaways

• Panther Lake platform enables 180 TOPS AI performance.
• iGPU supplies majority of AI compute, NPU adds efficiency.
• Same chassis as EVO T1, minimizing redesign costs.
• Configurations: 64‑96 GB RAM, 1‑2 TB SSD storage.
• Market impact hinges on price, availability, and support.

Summary

GMKtec unveiled the EVO T2 mini PC, claiming up to 180 AI TOPS powered by Intel’s Panther Lake Core Ultra X9 388H. The platform combines a Xe3‑based Arc iGPU (~122 TOPS), NPU5 (~50 TOPS) and CPU (~9 TOPS) to reach the figure. The design reuses the EVO T1 chassis and offers 64‑96 GB RAM with 1‑2 TB SSD options. Pricing and real‑world performance remain unconfirmed, but the device targets edge AI, local LLM inference, and media workloads.

Pulse Analysis

The AI edge market is accelerating as enterprises seek to process data locally, reducing latency and cloud costs. GMKtec’s EVO T2 aims to meet that demand by packing Intel’s 2026 Panther Lake SoC into a compact mini‑PC, promising up to 180 trillion operations per second (TOPS) of AI inference. This figure represents a substantial jump from the company’s previous EVO T1, which delivered roughly 99 TOPS, and positions the device alongside entry‑level desktop GPUs in raw AI throughput. If the claim holds, the T2 could become a reference point for high‑performance edge compute.

Panther Lake achieves the headline number through a tightly integrated trio of accelerators. The Xe3‑based Arc iGPU contributes around 122 TOPS, handling dense matrix operations typical of vision and language models. Complementing it, Intel’s NPU5 adds roughly 50 TOPS of dedicated inference power while keeping energy consumption low. The remaining ~9 TOPS come from the Core Ultra X9 388H CPU, which manages control flow and mixed‑precision tasks. Compared with AMD’s Ryzen AI Max 395 (≈126 TOPS), the EVO T2’s balanced architecture offers both higher peak performance and a more versatile software stack.

Real‑world impact will depend on price, availability, and support infrastructure. Should GMKtec price the EVO T2 competitively—below a mid‑range desktop workstation with a discrete GPU—it could attract small‑to‑medium businesses deploying local large‑language‑model inference, video transcoding, or industrial vision pipelines. Conversely, a premium price tag would relegate the unit to early‑adopter and niche markets. The reuse of the EVO T1 chassis lowers development risk, but thermal and power budgets remain critical constraints. Analysts will watch the launch closely as a barometer for how quickly AI‑centric silicon can migrate from data centers to the desktop.