LFM2
New generation of hybrid models for on-device edge AI
183 followers
New generation of hybrid models for on-device edge AI
183 followers
LFM2 by Liquid AI is a new class of open foundation models designed for on-device speed and efficiency. Its hybrid architecture delivers 2x faster CPU performance than Qwen3 and SOTA results in a tiny footprint.
This is the 5th launch from LFM2. View more
LFM2.5
Launched this week
LFM2.5 model family is Liquid AI's most capable release yet for edge AI deployment. It builds on the LFM2 device-optimized architecture and represents a significant leap forward in building reliable agents on the edge.
Free
Launch Team
Flowtica Scribe
Hi everyone!
Been following Liquid AI for quite a while, and their unwavering commitment to on-device models has always been impressive. Seeing them launch LFM2.5 alongside AMD at CES feels like a definitive milestone, it perfectly integrates into the new wave of AI PCs.
Fitting a full modal stack (Text, Vision, Audio) into the 1B parameter range is a smart move for edge constraints. The 8x speedup in the Audio model is a significant improvement for latency, and the specific optimizations for AMD and Qualcomm NPUs show that this is built for actual hardware.
I really think 2026 is going to be the year on-device AI finally scales up.
It's great to see on-device AI models. What are the minimum RAM requirements for LFM 2.5, and is it possible to run quantized versions?
Any idea how well this would run on a phone? Would love to try it without needing a full laptop setup.
Impressive direction. On-device speed + efficiency is where real adoption happens, especially for privacy-sensitive and latency-critical use cases. The hybrid architecture angle is interesting — curious to see how LFM2 performs in real-world edge scenarios compared to current lightweight LLMs.
Great project! I’m still waiting for models for regular phones that can work offline.