Introducing LFM2-VL — our new generation of efficient vision-language models for real-world deployment, from smartphones and laptops to wearables and embedded systems. 🧵

You can now fine-tune LFM2 models in Leap and directly deploy them in your application. instructions:

We’ve partnered with @dottxtai and @remilouf to make function calling viable on the edge. JSON doesn’t cut it. We need structure, speed, and reliability.

Try LFM2 with llama.cpp today! We released today a collection of GGUF checkpoints for developers to run LFM2 everywhere with llama.cpp Select the most relevant precision for your use case and start building today.

📲 @LiquidAI_ and Abu Dhabi’s @G42ai have partnered to support generative AI solutions driven by Liquid Foundation Models. The partnership aims to provide the Global South access to generative AI across sectors such as consumer electronics and biotech.

@LiquidAI_

Excited to share insights into our work on efficient reasoning @LiquidAI_ : We turn a 1.3B general chat model into a strong math reasoner that remains concise. SFT on curated CoT traces unlocks strong reasoning and targeted RL controls verbosity. The recipe is simple and provides an interesting datapoint for non-Qwen based models. This is still based on our first generation of LFMs - excited about the next generation! Please find details in our blog-post!

We are proud to announce @jonmcneill CEO of DVx, former president of @Tesla, former COO of @lyft, board member of @GM and @lululemon, joined us as a strategic member of our Advisory Council. Jon will help us fine-tune our growth strategy to bring private, on-device foundation models to market faster. He will guide us as we test our product market fit and connect with enterprise customers! welcome on board Jon McNeill!