Zant – A TinyML SDK in Zig
Hey HN, We’re excited to announce Zant v0.1, an open-source TinyML SDK built in Zig, designed to optimize and deploy neural networks on resource-constrained devices. Unlike existing solutions, Zant focuses on performance, portability, and ease of integration, making it a strong alternative for anyone working on Edge AI and embedded ML. Why Zant? Most TinyML frameworks are either too high-level (requiring bloated runtimes) or too low-level (requiring extensive manual optimization). Zant bridges the gap by offering: - A lightweight but powerful code generation system to translate ML models…
What it does
In the maker’s words, at launch
Hey HN, We’re excited to announce Zant v0.1, an open-source TinyML SDK built in Zig, designed to optimize and deploy neural networks on resource-constrained devices. Unlike existing solutions, Zant focuses on performance, portability, and ease of integration, making it a strong alternative for anyone working on Edge AI and embedded ML. Why Zant? Most TinyML frameworks are either too high-level (requiring bloated runtimes) or too low-level (requiring extensive manual optimization). Zant bridges the gap by offering: - A lightweight but powerful code generation system to translate ML models into optimized C/Zig code. - Better memory efficiency than Python-based tools like TensorFlow Lite Micro. - No runtime overhead—all computations are optimized for the target hardware. - A modern, memory-safe approach using Zig instead of C/C++. Features: - Code generation now supports 29 operations, including: GEMM (General Matrix Multiplication), Convolutions (Conv2D), Activation functions like ReLU, Sigmoid, Leaky ReLU, and more - Over 150 tests ensuring correctness and robustness across different hardware targets - A fuzzing system helps detect mathematical errors and verify the integrity of auto-generated code. Zant supports fully connected networks and simple convolutional architectures, making it suitable for various real-world TinyML applications. Supported Hardware: Zant has already been tested on multiple embedded platforms, showing promising results in real-world deployment: Raspberry Pi Pico (1 & 2) STM32 G4 and H7 Arduino Giga Seeed Camera More devices are being added as testing expands. Roadmap: Zant is still in early development, but we have ambitious goals for the next versions: Expanding code generation to cover more ML operations. - Quantization support (already in progress) to reduce model size and improve efficiency. - YOLO support for real-time object detection on microcontrollers. - Simplified deployment workflows to make it easier to use Zant across different hardware platforms. - CI/CD pipeline to improve reliability and automate testing. - Community engagement with a Telegram/Discord channel launching soon. Why Zig? Zig provides a modern, safer alternative to C, with better memory safety and performance optimizations. Unlike Python-based ML tools, Zant’s Zig-based approach avoids runtime overhead, making it ideal for low-power embedded devices. How to Get Involved: If you’re interested in TinyML, Edge AI, or embedded development, we’d love your feedback and contributions! No prior experience with Zig or TinyML is required—just a willingness to learn and a passion for the project. GitHub: https://github.com/ZantFoundation/Z-Ant Contributor Form:https://airtable.com/appYbTCd8vgMzJzFL/shrcWtM08l3VhAPM7 What do you think? What would you like to see next?
Does the same job
all alternatives →
- NONous – Open-Source Agent Framework with Autonomous, SWE Agents, WebUI2024 · github.com · ▲155
Hello HN! The day has finally come to stop adding features and start sharing what I've been building the last 5-6 months. It's a bit of CrewAI, OpenDevon, LangFuse/Cloud all in one, providing devs who prefer TypeScript an integrated framework thats provides a lot out of the box to start experimenting and building agents with. It started after peeking at the LangChain docs a few times and never liking the example code. I began experimenting with automating a simple Jira request from the engineering team to add an index to one of our Google Spanner databases (for context I'm the…
- LBLowdefy – Build internal tools with YAML on an open-source framework2021 · lowdefy.com · ▲107
- ZMZig monthly: iOS support, tutorials, tree-sitter, and more2021 · zigmonthly.org · ▲33
- ZEZenML, extensible open-source MLOps framework for reproducible ML2020 · github.com · ▲37
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 26d ago · cactuscompute.com
Launched alongside, March 2025
the whole month →
Mimic Human Research & Save Findings in AI Knowledge Base
AI · 2025 · sider.ai
