Documentation
Complete API reference & guides
Effortlessly ship powerful and performant AI in your apps — discover, specialize, and deploy models on any device, in minutes.
Endless possibilities with LEAP
How it works
1
Discover the best model for your use case within your constraints.
FEATURES
LFM2 is a hybrid-architecture class of Liquid Foundation Models (LFMs) that sets a new standard in quality, speed, and memory-efficient deployment. LFM2 is specifically designed to provide the fastest on-device gen-AI experience across the industry, thus unlocking a massive number of devices for generative AI workloads.
Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models.
LFM2 is a hybrid-architecture class of Liquid Foundation Models (LFMs) that sets a new standard in quality, speed, and memory-efficient deployment. LFM2 is specifically designed to provide the fastest on-device gen-AI experience across the industry, thus unlocking a massive number of devices for generative AI workloads.
LFM2 is a hybrid-architecture class of Liquid Foundation Models (LFMs) that sets a new standard in quality, speed, and memory-efficient deployment. LFM2 is specifically designed to provide the fastest on-device gen-AI experience across the industry, thus unlocking a massive number of devices for generative AI workloads.
3
Specialize models to your task with LEAP fine-tuning CLI and easily generate a deployment-ready model bundle.
FEATURES
Integrate your model seamlessly into your app, and run it entirely on-device.
FEATURES
1import ai.leap.LeapClient
2
3lifecycleScope.launch {
4 // Load the model
5 run = LeapClient.loadModel(BUNDLE_PATH)
6
7 // Create conversation
8 val conv = run.createConversation(SYSTEM_PROMPT)
9
10 // Run generation
11 conv.generateResponse(USER_PROMPT)
12 .onEach { chunk ->
13 when (chunk) {
14 is MessageResponse.Chunk -> {
15 val newContent = chunk.text
16 // process generated text ...
17 }
18 else -> {
19 // process other responses
20 }
21 }
22 }
23 .onCompletion {
24 // to perform when generation is done
25 }
26 .catch {
27 // error handling
28 }
29 .collect()
30}
Join the Liquid AI Hackathons to build cutting-edge on-device AI apps using LFM2 and LEAP, meet other innovative builders from around the world, and win prizes.
Up to 3 people per team
Dedicated Discord Community
Resources