Liquid AI's New 24B Model Runs on Your Laptop, Keeps Data Private

Liquid AI just launched a new AI model called LFM2.5-24B-A2B. It's a 'reasoning' model, meaning it thinks step-by-step before answering, like a person pondering a question. But unlike most powerful AI that lives in huge data centers, this one is built to run directly on your laptop or phone. That means your data stays private and you don't need an internet connection.

The model has 24 billion total parameters (the dials that control its behavior), but only 2 billion are active at any time. This clever design makes it fast and cheap to run. For example, on an Apple M5 Max chip, it can generate 184 tokens per second, which is faster than you can read. It also scores well on agent tasks, like using tools and following instructions, beating a larger model from Qwen.

Why should you care? If you use AI for work or personal tasks, this could mean a future where your AI assistant is always available, even offline, and your sensitive information never leaves your device. It supports many languages, including English, Spanish, Chinese, and Arabic. The model is free to try on Liquid AI's playground, and the code is available for download.

The catch: It's text-only for now, so no images or audio. Also, while it's efficient, running a 24B model on a device still requires a powerful computer. But it's a big step toward making advanced AI accessible and private for everyone.