MSI introduces the PRO MAX EDGE AI+, a 4-liter desktop delivering up to 126 TOPS of AI performance. The device runs on an AMD Ryzen AI Max+ 395 processor with up to 128 GB of shared memory and is intended for professionals who want to execute AI workloads locally without relying on the cloud.
With the PRO MAX EDGE AI+, MSI targets developers, researchers, and small studios that want to maintain control over their data. By performing AI processing on the device itself, sensitive data does not need to leave the network, which benefits responsiveness and data sovereignty.
Powerful hardware in a small chassis #
The heart of the system is the AMD Ryzen AI Max+ 300 series, with the Ryzen AI Max+ 395 as the flagship model. That chip combines sixteen Zen 5 cores, forty RDNA 3.5 compute units, and an XDNA 2 NPU that reaches up to 50 TOPS. Together, the CPU, GPU, and NPU deliver up to 126 TOPS of total AI performance.
The device supports up to 128 GB of LPDDR5X memory at 8,000 MT/s, of which up to 96 GB can be allocated as graphics memory. Because the CPU and integrated GPU share this memory, compatible applications can run large language models with up to 120 billion parameters locally. Results depend on model architecture, quantization, and system configuration.
Clustering for larger models #
Teams with heavier workloads can combine multiple PRO MAX EDGE AI+ systems into a local cluster. According to MSI’s validation table, two nodes are sufficient for certain models with 120 and 235 billion parameters, while a four-node setup can handle DeepSeek-V3.1 with 671 billion parameters at Q4_0 quantization.
For cooling, MSI relies on Frozr AI Pro and Glacier Armor, featuring a copper heat spreader, three heat pipes, and three fans. The four-liter aluminum chassis integrates a 300 W power supply and offers, among other features, Wi-Fi 7, four independent display outputs, and two 40 Gbps USB ports.
Compact solution for local AI #
With the PRO MAX EDGE AI+, MSI is responding to the demand for compact workstations capable of performing AI inference, RAG applications, and data analysis locally. Prices and availability for specific configurations have not yet been announced.