Notes:
Expected to ship from Hong Kong around end-August.
*Tested with Intel® Core™ Ultra 9 386H, 64GB memory, llama-b8783-win-vulkan-x64, and Qwen3.5-35B-A3B-Q4-K-M.
Qwen 35B is supported with 48GB or more of system memory.
U.2 enterprise SSD support via adapter gives large model files, vector databases, and high-read workloads a stronger local storage foundation.
Three M.2 SSD slots—two PCIe 5.0 and one PCIe 4.0—let you separate the OS, models, and datasets for cleaner local AI workflows and easier upgrades.
Supports low-profile add-in cards for capture, networking, and specialized deployments.
Dual heatpipes rapidly move heat away from the core, helping reduce performance fluctuations under sustained workloads.
Dual heatpipes rapidly move heat away from the core, helping reduce performance fluctuations under sustained workloads.
Focused airflow exhausts heat efficiently, helping the compact chassis sustain performance.
Focused airflow exhausts heat efficiently, helping the compact chassis sustain performance.
Ready for local model inference, edge services, knowledge-base search, and 24/7 workloads.
Ready for local model inference, edge services, knowledge-base search, and 24/7 workloads.
Dual heatpipes rapidly move heat away from the core, helping reduce performance fluctuations under sustained workloads.
Dual heatpipes rapidly move heat away from the core, helping reduce performance fluctuations under sustained workloads.
Focused airflow exhausts heat efficiently, helping the compact chassis sustain performance.
Focused airflow exhausts heat efficiently, helping the compact chassis sustain performance.
Ready for local model inference, edge services, knowledge-base search, and 24/7 workloads.
Ready for local model inference, edge services, knowledge-base search, and 24/7 workloads.