At Pi Australia, we’re excited to announce the arrival of the Raspberry Pi AI HAT+ 2 — the next-generation AI accelerator that brings local, generative AI to the Raspberry Pi 5 for the very first time.
Designed for developers, educators, and innovators, the AI HAT+ 2 unlocks powerful on-device AI capabilities without the cost, latency, or privacy concerns of cloud-based services. If you’ve been waiting to run modern AI workloads directly on Raspberry Pi hardware, this is the upgrade you’ve been looking for.
👉 Available now in Australia: https://piaustralia.com.au/products/raspberry-pi-ai-hat-plus-2
What’s New with AI HAT+ 2?
The original AI HAT+ introduced high-performance edge AI for computer vision. The AI HAT+ 2 takes the next leap forward by enabling generative AI workloads, including Large Language Models (LLMs) and Vision-Language Models (VLMs) — all running locally on a Raspberry Pi 5.
At the heart of the board is the Hailo-10H neural network accelerator, delivering up to 40 TOPS (INT4) of inferencing performance. Combined with 8GB of onboard LPDDR4X memory, the AI HAT+ 2 can efficiently handle larger and more complex AI models than ever before.
All processing happens directly on the device, ensuring:
- Low latency
- No internet dependency
- Improved data privacy
- Predictable, cloud-free operating costs
Generative AI at the Edge — No Cloud Required
With AI HAT+ 2, Raspberry Pi users can now run compact but capable LLMs locally. Supported models at launch include:
- Llama 3.2
- Qwen 2 / Qwen 2.5 (including Coder variants)
- Vision-language models for camera-based understanding
- Whisper (base) for speech-to-text
These models enable use cases such as:
- Local chatbots and coding assistants
- Text translation and summarisation
- Voice interfaces and transcription
- Camera-based scene description and analysis
For Australian educators, researchers, and businesses, this means AI experimentation and deployment without sending sensitive data offshore.
Still Excellent for Computer Vision
If you’re already using the original AI HAT+ for vision workloads, the good news is that the AI HAT+ 2 maintains comparable performance for computer vision tasks such as:
- Object detection (YOLO-based models)
- Pose estimation
- Scene segmentation
Thanks to the onboard RAM and tight integration with the Raspberry Pi camera software stack (libcamera, rpicam-apps, Picamera2), upgrading from AI HAT+ to AI HAT+ 2 is largely seamless.
Built for Developers, Educators, and Industry
The AI HAT+ 2 is ideal for:
- Edge AI and robotics projects
- Smart cameras and monitoring systems
- AI education and research
- Prototyping privacy-first AI solutions
- Local AI assistants and tools
It also supports model fine-tuning, including:
- Retraining vision models using the Hailo Dataflow Compiler
- LoRA-based fine-tuning of language models for task-specific AI
Available Now from Pi Australia 🇦🇺
The Raspberry Pi AI HAT+ 2 is available now from Pi Australia, with local stock and Australian support.
🔗 Buy the Raspberry Pi AI HAT+ 2 here: https://piaustralia.com.au/products/raspberry-pi-ai-hat-plus-2
If you’re looking to bring powerful, private, and affordable AI to your Raspberry Pi 5 projects, the AI HAT+ 2 is the most exciting Raspberry Pi accessory we’ve seen in years.
As always, feel free to contact the Pi Australia team if you need help choosing accessories, setting up your board, or exploring AI use cases.
