LattePanda Mu Ultra - A Micro x86 Compute Module for On-device AI (Intel Core Ultra 5 226V CPU, 16GB RAM, 97 TOPS)

LattePanda Mu Ultra 226V is a micro x86 compute module built for on-device AI. Powered by an Intel Core Ultra 5 226V processor (Lunar Lake) with 16GB of LPDDR5X memory running at 8533MT/s, it delivers up to 97 TOPS of combined AI performance across its CPU, GPU, and NPU—all on a credit-card-sized board. With rich expansion pins (PCIe 4.0, USB 3.2 Gen 2, HDMI/DP outputs, etc.) and multi-OS support, it makes it suitable to integrate powerful local AI into robots, handhelds, and edge devices.

Massive AI Power, Up to 97 TOPS for Local AI

The LattePanda Mu Ultra 226V processor packs three compute engines into one module: an 8-core CPU, an Intel Arc GPU, and an Intel AI Boost NPU — delivering up to 97 TOPS of heterogeneous AI compute in total.

The CPU, GPU, and NPU each handle the workloads they’re best suited for, enabling local execution of LLMs, VLMs, YOLO-based vision models, and speech models without a cloud connection, with no cloud connection required.

Powerful General Performance

LattePanda Mu Ultra is also a seriously capable computer-on-module. The 8-core CPU and the Arc GPU enable logic control, accelerated graphics, and parallel computing with desktop-class performance. This allows the LattePanda Mu Ultra to handle dense workloads with responsive performance.

At a 37W TDP setting, the LattePanda Mu Ultra 226V achieved approximately 9,800 multi-core and 2,400 single-core points in Geekbench 6. Its Intel Arc 130V GPU scored approximately 2,900 in the 3DMark Time Spy Graphics test. In OpenVINO NPU forward-pass testing, the module delivered 235.3 FPS with YOLO26n, 158.7 FPS with YOLO26s and 78.2 FPS with YOLO26m.

Lightning-Fast Unified Memory On Package

Local LLM performance relies heavily on memory bandwidth. The LattePanda Mu Ultra features a Memory-on-Package (MoP) with 16GB of LPDDR5X at 8533 MT/s, delivering up to 136.5GB/s of bandwidth for significantly faster token generation.

In addition, allocate up to 11.6GB as shared GPU memory to run larger models or support expanded context windows.

Card-Sized, Fits Anywhere

LattePanda Mu Ultra measures only 69.6mm x 60mm. The credit-card-sized form factor concentrates the processor, 16GB memory and expansion signals into a compact module suitable for space-constrained systems such as mobile robots, handheld instruments and smart terminals.

Active or Idle, Efficient Power Consumption for Edge AI

LattePanda Mu Ultra delivers more performance per watt than previous generations—so every token, frame, and inference consumes less energy.

Low idle power consumption also suits local AI workloads, which often consist of short inference bursts followed by longer idle periods. Efficient power consumption under both active and idle conditions helps extend operating time in mains-powered and battery-powered applications.

Support for Popular AI Software Frameworks

LattePanda Mu Ultra supports widely used AI software frameworks and runtimes. Intel OpenVINO enables acceleration across the CPU, GPU, and NPU, while runtimes and frameworks such as Ollama can run directly on the platform. Support for LLMs, VLMs, YOLO, and speech models helps simplify development and migration with familiar software tools.

Multi OS Support

The LattePanda Mu Ultra x86 AI compute module supports multiple operating systems, including Windows 11 and Ubuntu 24.04 or later, providing options for different development and deployment requirements.

Flexible Expansion Pins

LattePanda Mu Ultra provides extensive pin access, including flexible PCIe 4.0 configurations with up to 4 x1, 4 x2, or 2 x4 links, two USB 3.2 Gen 2 ports, six USB 2.0 ports, three HDMI/DP outputs, and additional expansion signals. This flexibility supports the development of application-specific hardware configurations.

Ready-to-Use Carrier Board for Faster Development

As a compute module, LattePanda Mu Ultra requires a compatible carrier board to provide power input, storage expansion and physical peripheral connectors.

The Mini Carrier Board offers a compact development platform with two full-featured USB Type-C ports, two USB 3.2 Gen 2 Type-A ports, 2.5G Ethernet, HDMI 2.0, OCuLink and separate M.2 slots for NVMe SSD and WLAN expansion.

The Lite Carrier Board also supports LattePanda Mu Ultra, with limitations on some ports. See the compatibility list for details before selecting interfaces for a Mu Ultra configuration.

Open-Source Carrier Board Reference Files

LattePanda provides open-source carrier board reference files for adapting carrier board designs to application-specific interface, layout and form-factor requirements. These reference materials help reduce the work required to begin a custom carrier board design.

Customized Solutions

LattePanda offers customization services for carrier boards, boot screens, BIOS functions, operating systems, and related requirements. Specific customization requests can be sent to [email protected].

LattePanda provides technical support for the evaluation and implementation of customized solutions.

Note: The LattePanda Mu Ultra series includes onboard memory but no onboard storage. A compatible SSD, which is not included, must be installed through the carrier board’s M.2 slot before use.