Liquid AI's Tiny Decision Models Bring Fast AI to Your Phone

Liquid AI just launched two new AI models called d1-3B and d1-omni-600M. Unlike ChatGPT-style tools that generate text, these models are designed to make quick decisions, like sorting emails or recognizing objects in photos. They're small enough to run directly on your phone or laptop, not in a distant data center. That means faster responses and better privacy, since your data doesn't leave your device.

These models are built on Liquid's foundation models and can handle text, images, and audio. The larger d1-3B scored highest on a set of decision-making tests, beating similar models. The smaller d1-omni-600M is super compact but still performs well, making it ideal for devices with limited power. In tests with NVIDIA hardware, d1-3B answered questions in under 50 milliseconds on devices like the Jetson Orin Nano, and even faster on GPUs.

So what does this mean for you? Imagine a fitness app that instantly analyzes your form through your phone's camera, or a messaging app that filters spam without sending your messages to a server. These models make that possible. They're open-weight, meaning developers can download and use them for free from Hugging Face. This could lead to a wave of new, responsive apps that work offline and respect your privacy.

The catch: these aren't general-purpose AI like ChatGPT. They're specialized for specific decision tasks, so you won't be having conversations with them. Also, the smallest model is still experimental, so it might not be ready for prime time. But for developers, it's a big step toward powerful AI that runs right in your pocket.