Liquid AI

Proud of the team and our release of LFM2.5-8B-A1B for on-device agentic tasks! Since the LFM2-8B-A1B release, we expanded pre-training, increased the vocab for improved efficiency on non-Latin scripts, extended context length, and made significant post-training advances. HF: https://lnkd.in/gJDkGqdS HF GGUF: https://lnkd.in/gm6AgrJk Blog: https://lnkd.in/giff4sdU

Today, we're releasing LFM2.5-8B-A1B, an edge model designed to power real-life applications. It builds on LFM2-8B-A1B with three major upgrades: an expanded 128K context window, 38T tokens of pre-training (up from 12T), and large-scale reinforcement learning. We also doubled the vocabulary to improve tokenization for non-Latin languages. The result is a model that chains tool calls, completes complex tasks, and fits comfortably on an entry-level laptop. Most on-device models are built for chat. LFM2.5-8B-A1B is built for agents: Ask, propose, confirm, run, repeat. All in well under a second per dispatch, with your data never leaving the device. To show what fast, reliable tool calling looks like in practice, we ran it on LocalCowork, our open-source desktop agent: A single laptop, 67 tools across 13 MCP servers, no cloud, no API keys. Watch the 3-minute demo → https://lnkd.in/ewwsKH-C From day one, LFM2.5 runs with native support for llama.cpp, MLX, vLLM, SGLang across Apple, AMD, Intel, Qualcomm, and NVIDIA hardware. Start building today with LFM2.5-8B-A1B, available on Hugging Face, LEAP, and our Playground. Read more: https://lnkd.in/etHZdQ2C

The Liquid AI × DPhi Space “AI in Space” hackathon asked builders: What becomes possible when state-of-the-art models run in orbit? We’re excited to announce the winning teams that answered it best: GalamseyWatch, by Samuel Donkor, of Ghana. And Parali, by Munish Kumar and Aashish Kumar, of India. Thank you to DPhi Space for co-hosting, and to everyone who participated!

The demand for AI is growing exponentially 📈 But the energy to power it isn’t. Our CEO Ramin Hasani sat down for a conversation with CTO Mathias Lechner to discuss how Liquid AI is working to solve a core issue as demand for compute skyrockets.

Well said, Jeffrey Li. AI shouldn't just live in a data center, but it should run on the device you're already using: Whether it's a phone, a car, an appliance, or another everyday device in the real world. That’s what we’re working on at Liquid AI, and we appreciate the spotlight on our COO on the AI-Curious Podcast. Thank you, Jeff Wilser! Listen to the full episode at the link in comments.

We’re coming to Tokyo for a hackathon, along with WAY Equity Partners and AMD ✨ Join us there from June 6-7. Over two days, you’ll collaborate with engineers, founders, and mentors across the Liquid AI and WAY ecosystems to use our Liquid Foundation Models (LFMs). We’ll reveal the challenge on-site. Teams of 1-3 will receive their share of the Gold $3K USD prize or Silver $2K USD prize. Apply to join: https://luma.com/7fjlam5k Start building with LEAP: leap.liquid.ai

LFM2-VL-3B in space ✨ Big milestone for Liquid AI and DPhi Space, as our vision-language model runs on Clustergate-2: When our model was running aboard a satellite in orbit, it was prompted to describe an image and said: "This image is a highly detailed, close-up view of Earth as seen from space, likely captured by a satellite or space telescope. The Earth is depicted as a large, circular sphere with a predominantly blue hue, indicating the vast oceans that cover most of its surface. The blue is interspersed with swirling white clouds, which are particularly prominent over the landmasses, suggesting the presence of weather systems and atmospheric activity. The overall composition of the image highlights the beauty and complexity of our planet, showcasing the dynamic interplay between the oceans, atmosphere, and landmasses." Orbital infrastructure is becoming programmable, accessible, and intelligent, and we’re proud to partner with DPhiSpace on this progress! Read more, from the first public glimpse of Clustergate-2: https://lnkd.in/eej4MnAQ

⚡ 10x Smaller, 10x Faster: New SOTA Benchmarks for Pharma #LLMs Most LLMs fail and continue to miss the mark on critical ADMET and toxicity tasks. Domain expertise isn't solved by prompt tuning alone, it requires specialized training. Through our partnership with Liquid AI, we’ve utilized the MMAI Gym for Science to transform the LFM2-2.6B model into a #drugdiscovery powerhouse. Despite its small footprint, it is now delivering State-of-the-Art (SOTA) results that rival or surpass models ten times its size. 💥 Key Benchmarks at a Glance: -ADMET Supremacy: Surpassed TxGemma-27B across multiple property prediction endpoints. -94% #Retrosynthesis Accuracy: Achieved via specialized SFT+RFT training sessions. -Undisputed SOTA: Leading performance in functional group reasoning (FGBench) and molecular optimization (MuMO). -Edge-Ready: Optimized for CPUs, NPUs, and GPUs, providing low-latency, "always-on" intelligence without the need for cloud calls. We’re ready to help you integrate these models directly into your DMTA cycle for real-time lead optimization. See it in action: Test the MMAI Demo here 👉 https://lnkd.in/eDtrb7UE Ready to deploy? Contact our team at bd@insilicomedicine.com #MMAIGym

Start your engines. The news about our partnership with Mercedes-Benz AG is lighting up Times Square today. Thank you to Brian J. Baumann and the NYSE for the support! Read more about the partnership, announced today: https://lnkd.in/gJCRQ5mh

I’m proud to announce our new partnership with Mercedes-Benz AG to bring embedded, on-device intelligence to Mercedes-Benz vehicles starting in North America vehicles. This marks an important step toward making in-car AI more capable, more responsive, and more useful in everyday driving. At Liquid AI, we believe the future of intelligence in the physical world depends on models that are fast, private, efficient, and able to run directly on the hardware already inside the system. In the vehicle, that means advancing speech, language understanding, and reasoning enabling more natural and robust conversational experiences for drivers and passengers. The software-defined vehicle is one of the most consequential real-world deployments of AI, and Mercedes-Benz has approached it with exactly the rigor this challenge deserves. Proud of what our teams are building together, and excited for the road ahead as we work toward production deployment in the second half of 2026. I am also deeply grateful of the support of the Mercedes-Benz executives, Joerg Burzer, Magnus Östberg, and Jason Hoff to kick off this innitiative, with many impactful future prospects. Press: https://lnkd.in/gQacCDw4 Blog: https://lnkd.in/gniyZRzN