LFM3 by Liquid AI is the third generation of Liquid Foundation Models, released with open weights and a shared backbone spanning text, vision, audio, omni, and decision models. The initial release includes six 2B-class models, among them a 1.7B text model that fits in roughly 1 GB at 4-bit precision and supports a 256K context. The models are designed to run on phones, laptops, and edge devices and can be fine-tuned on a single GPU. Weights are available on Hugging Face.
LFM3 by Liquid AI
Six 2B-class models for phones, laptops and robots
LFM3 by Liquid AI Introduction
Key Features
- Third generation of Liquid Foundation Models with open weights
- Shared backbone architecture across text, vision, audio, omni, and decision models
- Six 2B-class models in the first release
- 1.7B text model that fits in about 1 GB at 4-bit
- 256K context length on the compact text model
- Runs on phones, laptops, and edge devices
- Fine-tunable on a single GPU
- Weights published on Hugging Face
Use Cases
- Deploying compact language models on mobile devices and edge hardware
- Fine-tuning small foundation models on limited GPU resources
- Building multimodal applications with a unified model family
- Integrating open-weight LLMs into offline or on-device products
Why Startups Use It
LFM3 gives startups open-weight, compact foundation models that can run directly on consumer and edge hardware, making it possible to build AI-powered features without relying on large infrastructure. A 1.7B text model that fits in about 1 GB at 4-bit and supports a 256K context, combined with single-GPU fine-tuning, lowers the cost and complexity of shipping on-device AI experiences.
Alternative Options
Llama, Mistral, Gemma, Phi, Qwen
Frequently Asked Questions
What is LFM3?
LFM3 is Liquid AI's third generation of Liquid Foundation Models, offered with open weights and a shared backbone across text, vision, audio, omni, and decision models.
How large are the models?
The first release consists of six 2B-class models, including a 1.7B text model that fits in about 1 GB at 4-bit precision.
What context length does the text model support?
The 1.7B text model supports a 256K context.
Where can the weights be accessed?
The model weights are available on Hugging Face.
What hardware can LFM3 run on?
Liquid says the models run on phones, laptops, and edge devices, and can be fine-tuned on one GPU.
More About LFM3 by Liquid AI
PricingPaid
Listed DateOct 29, 2026
Authority Badge
Add our badge to your website to showcase product credibility and listing status.
Featured List