Raspberry Pi Features New Sixfab AI Accelerator
The new Sixfab AI HAT+ board leverages DeepX technology to bring high-performance, on-device artificial intelligence to users.
Updated on Oct. 15, 2026 in Artificial Intelligence
Live Poll
Would you use specialized AI hardware to upgrade your existing home computing devices?
Raspberry Pi has highlighted a new AI acceleration board developed by Sixfab that utilizes the DX-M1M neural processing unit. This integration allows for robust on-device AI tasks, removing the need to transmit data to external servers.
Why it matters
On-device processing provides users with enhanced data privacy and operational efficiency by eliminating the requirement to send information to cloud servers. This development marks a significant shift in the accessibility of local AI compute power for developers.
The DX-M1M NPU features 2GB of LPDDR4X memory and delivers 25 TOPS of performance while consuming 3W during continuous use. It achieves processing speeds of 2,361 FPS for MobileNet V2 and 155 FPS for DeepLab V3+.
The players
Raspberry Pi
This organization develops affordable, high-performance single-board computers widely used by students and engineers.
Sixfab
The company specializes in modular hardware add-ons that expand the connectivity and computing capabilities of small computers.
DeepX
This firm focuses on the design of efficient neural processing units and specialized AI software stacks for hardware integration.
The details
Developers utilize DX-COM for model conversion, DX-RT for inference software, and DX-Stream for video processing tasks. The NPU also generates text at 4.64 tokens per second with a 599-millisecond initial response latency when running the Qwen3 1.7B model.
Timeline
June 2026: DeepX and Sixfab released the AI acceleration board.
October 5, 2026: Raspberry Pi published the official introductory article.
October 15, 2026: DeepX announced the Raspberry Pi publication.
Later in 2026: The 13 TOPS version of the product is scheduled for release.
The Tech Race
The introduction of specialized AI accelerators marks a shift away from relying on cloud-based processing toward powerful, efficient edge computing. This positions small-form-factor devices to compete with larger hardware setups in local machine learning tasks.
Users can now execute complex machine learning models directly on their hardware without requiring a constant internet connection. This improves latency and provides a practical way for developers to build private AI applications at home.
The takeaway
Moving AI workloads to local devices enables faster, private model execution without bandwidth limitations. Developers looking to implement AI should evaluate their specific TOPS requirements against power consumption constraints before selecting their hardware.
What happens next
A secondary, lower-power version of the AI acceleration board, scaled to 13 TOPS, is scheduled for launch later in 2026.
Further reading
For more developments in edge computing, explore the latest updates in Artificial Intelligence.
Source note: This article includes information reported by 조선일보.
Live Poll
Would you use specialized AI hardware to upgrade your existing home computing devices?
