Liquid AI has launched LFM2-VL, a brand new era of vision-language basis fashions designed for environment friendly deployment throughout a variety of {hardware} — from smartphones and laptops to wearables and embedded techniques.
The fashions promise low-latency efficiency, sturdy accuracy, and adaptability for real-world functions.
LFM2-VL builds on the corporate’s present LFM2 structure, extending it into multimodal processing that helps each textual content and picture inputs at variable resolutions.
In line with Liquid AI, the fashions ship as much as twice the GPU inference pace of comparable vision-language fashions, whereas sustaining aggressive efficiency on widespread benchmarks.
“Effectivity is our product,” wrote Liquid AI co-founder and CEO Ramin Hasani in a submit on X asserting the brand new mannequin household:
Two variants for various wants
The discharge contains two mannequin sizes:
- LFM2-VL-450M — a hyper-efficient mannequin with lower than half a billion parameters (inner settings) aimed toward extremely resource-constrained environments.
- LFM2-VL-1.6B — a extra succesful mannequin that is still light-weight sufficient for single-GPU and device-based deployment.
Each variants course of photographs at native resolutions as much as 512×512 pixels, avoiding distortion or pointless upscaling.
For bigger photographs, the system applies non-overlapping patching and provides a thumbnail for international context, enabling the mannequin to seize each superb element and the broader scene.
Background on Liquid AI
Liquid AI was based by former researchers from MIT’s Laptop Science and Synthetic Intelligence Laboratory (CSAIL) with the objective of constructing AI architectures that transfer past the broadly used transformer mannequin.
The corporate’s flagship innovation, the Liquid Basis Fashions (LFMs), are based mostly on rules from dynamical techniques, sign processing, and numerical linear algebra, producing general-purpose AI fashions able to dealing with textual content, video, audio, time sequence, and different sequential information.
In contrast to conventional architectures, Liquid’s strategy goals to ship aggressive or superior efficiency utilizing considerably fewer computational assets, permitting for real-time adaptability throughout inference whereas sustaining low reminiscence necessities. This makes LFMs properly suited to each large-scale enterprise use circumstances and resource-limited edge deployments.
In July 2025, the corporate expanded its platform technique with the launch of the Liquid Edge AI Platform (LEAP), a cross-platform SDK designed to make it simpler for builders to run small language fashions immediately on cell and embedded units.
LEAP presents OS-agnostic assist for iOS and Android, integration with each Liquid’s personal fashions and different open-source SLMs, and a built-in library with fashions as small as 300MB—sufficiently small for contemporary telephones with minimal RAM.
Its companion app, Apollo, allows builders to check fashions completely offline, aligning with Liquid AI’s emphasis on privacy-preserving, low-latency AI. Collectively, LEAP and Apollo mirror the corporate’s dedication to decentralizing AI execution, decreasing reliance on cloud infrastructure, and empowering builders to construct optimized, task-specific fashions for real-world environments.
Pace/high quality trade-offs and technical design
LFM2-VL makes use of a modular structure combining a language mannequin spine, a SigLIP2 NaFlex imaginative and prescient encoder, and a multimodal projector.
The projector features a two-layer MLP connector with pixel unshuffle, decreasing the variety of picture tokens and bettering throughput.
Customers can modify parameters similar to the utmost variety of picture tokens or patches, permitting them to stability pace and high quality relying on the deployment state of affairs. The coaching course of concerned roughly 100 billion multimodal tokens, sourced from open datasets and in-house artificial information.
Efficiency and benchmarks
The fashions obtain aggressive benchmark outcomes throughout a spread of vision-language evaluations. LFM2-VL-1.6B scores properly in RealWorldQA (65.23), InfoVQA (58.68), and OCRBench (742), and maintains stable leads to multimodal reasoning duties.
In inference testing, LFM2-VL achieved the quickest GPU processing occasions in its class when examined on an ordinary workload of a 1024×1024 picture and brief immediate.
Licensing and availability
LFM2-VL fashions can be found now on Hugging Face, together with instance fine-tuning code in Colab. They’re appropriate with Hugging Face transformers and TRL.
The fashions are launched below a customized “LFM1.0 license”. Liquid AI has described this license as based mostly on Apache 2.0 rules, however the full textual content has not but been printed.
The corporate has indicated that industrial use will likely be permitted below sure situations, with completely different phrases for firms above and beneath $10 million in annual income.
With LFM2-VL, Liquid AI goals to make high-performance multimodal AI extra accessible for on-device and resource-limited deployments, with out sacrificing functionality.
