Raspberry Pi

Raspberry Pi AI HAT+ 2

Overview

The Raspberry Pi AI HAT+ 2 is a generative AI accelerator add-on board for Raspberry Pi 5, released in January 2026. Featuring the Hailo-10H neural network accelerator with 40 TOPS (INT4) performance and 8GB of dedicated on-board LPDDR4X RAM, it enables large language models (LLMs), vision-language models (VLMs), and coding agents to run entirely locally on a Raspberry Pi 5 with no cloud dependency. The AI HAT+ 2 supports fine-tuning via LoRA adapters and integrates with the hailo-ollama backend and Open WebUI for a browser-based chat interface.

Key Highlights
  • First Raspberry Pi AI accelerator purpose-built for generative AI — runs LLMs, VLMs, and coding agents entirely on-device with no cloud dependency
  • Hailo-10H NPU delivers 40 TOPS (INT4) inferencing performance with 8GB dedicated LPDDR4X RAM, enabling 1-7 billion parameter models locally
  • Supports LLMs at launch including DeepSeek-R1-Distill 1.5B, Llama 3.2 1B, Qwen2.5-Coder 1.5B, and Qwen2.5-Instruct 1.5B via hailo-ollama backend
  • LoRA-based fine-tuning support via Hailo Dataflow Compiler for task-specific model customization
  • Backward compatible with existing AI HAT+ vision pipelines — YOLO, pose estimation, scene segmentation work seamlessly through Picamera2

Technical Specifications

Brand Raspberry Pi
Model Raspberry Pi AI HAT+ 2
Memory (RAM) 8GB LPDDR4X Onboard Dedicated AI Memory
Ports & I/O 1x 22-pin GPIO HAT Header (PCIe 2.0/3.0 x1)
1x FPC Connector for Camera
Connectivity PCIe via GPIO HAT Header (uses host Raspberry Pi 5 Wi-Fi/BT)
Operating System Add-on Board (requires Raspberry Pi 5 running Raspberry Pi OS)
Dimensions 66 x 56.5 x 21 mm