Under the lid sits AMD's Ryzen AI Max+ 395, a 16-core Zen 5 processor with an integrated Radeon 8060S GPU and a 50 TOPS NPU (a neural processing unit, dedicated to AI maths). The headline figure is up to 128GB of unified LPDDR5x-8000 memory, of which 96GB can be allocated to the GPU. MSI says that is enough to run large language models locally, including GPT-OSS 120B, which it lists as needing 65GB of VRAM.

The claimed numbers are MSI's own, so treat them as vendor figures rather than independent benchmarks. The company lists GPT-OSS 120B at an average 33.29 tokens per second, Qwen 3.6 35B at 58.96 and GPT-OSS 20B at 44.36. Beyond chatbots, MSI pitches the box at 3D rendering, high-resolution video editing and fully offline work such as image generation, coding and document analysis. Up to four units can also be clustered, depending on the model, for larger local workloads.

Connectivity covers four displays, two USB 4 40Gbps Type-C ports with DisplayPort Alt mode and 15W power delivery, USB 3.2 10Gbps ports, HDMI 2.1, DisplayPort 1.4a, a 2.5GbE LAN socket, an SDXC card reader and Wi-Fi 7 with Bluetooth 5.4. MSI also advertises 80fps-plus average frame rates at 2K with frame generation in AAA games, though it concedes that will vary by title.

The pitch is running models entirely offline rather than in the cloud, which MSI argues cuts operating costs and keeps sensitive business data off other people's servers. Quiet operation is credited to a pure copper heatsink and three fans.