Liquid AI’s LEAP Platform Brings Private, On-Device AI to the Edge—No Cloud Required
Liquid AI is cutting the cord on cloud dependence. The foundation model company today launched LEAP v0 (Liquid Edge-AI Platform), a new developer platform designed to bring advanced AI to smartphones, laptops, wearables, drones, and even cars—without relying on the cloud.
Also introduced is Apollo, an iOS-native app showcasing the potential of Liquid’s small foundation models (SLMs) in real-time, fully private interactions. The goal? Put the power of AI in your pocket—and keep your data out of someone else’s server farm.
Why It Matters: Edge AI Is the Next Frontier, But It’s Been Hard to Build
Cloud-based AI might power your ChatGPTs and Google assistants, but for privacy-conscious users and resource-limited environments, it’s a non-starter. The complexity of building AI that runs efficiently on edge devices has left many developers frustrated—or out of the game entirely.
LEAP aims to change that.
Built from scratch for developers—not just AI experts—LEAP allows the deployment of small foundation models (like Liquid’s new LFM2) in just 10 lines of code. That’s not hyperbole: it works cross-platform (Android and iOS) and doesn’t require advanced AI knowledge to get started.
“LEAP is our answer—a deployment platform designed from the ground up to make powerful, efficient, and private edge AI easy and accessible,” said Ramin Hasani, CEO and co-founder of Liquid AI.
Apollo: A Glimpse Into the Edge AI Experience
To show what edge-native AI can really do, Liquid is also launching Apollo, a sleek, iOS-native app that puts their models to work in a real-world context. Originally designed by indie developer Aaron Ng and now evolved by Liquid, Apollo offers a minimalist, intuitive interface for interacting with Liquid’s LFM2 models—all fully offline, with no cloud processing.
Apollo serves as both demo and testbed. Want to try out instruction-following models on your phone with zero latency and full data privacy? Apollo does that. Need something you can benchmark against other small language models? Apollo does that too.
And yes, Android support is on the way.
What’s Under the Hood: LFM2 and a New Architecture for SLMs
Both LEAP and Apollo are powered by Liquid AI’s LFM2 models, a new generation of open-source small foundation models that boast leading performance in the edge class. These aren’t just slimmed-down transformer models—they’re built from the ground up using structured, adaptive operators, an architecture Liquid claims offers:
- Faster inference
- Lower energy usage
- Improved generalization, especially in long-context or low-resource environments
LFM2 models are already making waves for their ability to outperform rival open-source SLMs in instruction-following and latency benchmarks—all while being light enough to run locally on consumer-grade hardware.
Developer Focus, Enterprise Implications
LEAP’s focus is refreshingly practical. Unlike edge solutions from Apple CoreML or Qualcomm AI Stack, which can be heavily hardware-dependent or require specialized knowledge, LEAP is platform-agnostic and developer-friendly, aimed at everyone from indie app devs to enterprise product teams.
Key features include:
- No cloud dependency: Fully offline deployment means data never leaves the device.
- 10-line setup: Start running AI in your app with minimal code.
- SLM library access: Built-in support for LFM2 and other edge-optimized models.
- Cross-platform support: Designed for iOS and Android integration out of the box.
The implications go far beyond consumer gadgets. In industries where privacy, latency, or connectivity are mission-critical—think healthcare, automotive, defense, and industrial IoT—LEAP could be a game changer.
The Bigger Picture: Reclaiming AI from the Cloud
With AI workloads increasingly concentrated in the hands of a few hyperscale cloud providers, Liquid AI’s approach feels both countercultural and timely. By prioritizing privacy, efficiency, and autonomy, LEAP and Apollo offer a compelling alternative for developers who want to build with AI—without the baggage of centralized control, API limits, or cloud costs.
It’s a bet that smaller, smarter, and local will win out in the long run—and Liquid AI may have just given edge AI the shot in the arm it’s been waiting for.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI.
