🔍 Read the full analysis: A Closer Look At Building Custom Multimodal Models With LFM3 on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Liquid AI has introduced LFM3, a model family built from a shared 1.7-billion-parameter base and adapted for text, vision, audio, multimodal inputs and decision tasks. The company says the models are designed for single-GPU fine-tuning and edge deployment, but the performance and efficiency figures provided are company-reported and have not been independently verified in the supplied material.
Liquid AI has introduced LFM3, a family of models built from a shared 1.7-billion-parameter base and adapted for text, vision, audio, multimodal inputs and decision tasks, as described in the original analysis. The company says developers can fine-tune the models on a single GPU and deploy them on edge devices or other infrastructure, but its performance and efficiency figures remain company-reported in the material available.
The family includes LFM3-1.7B for text, LFM3-VL-2B for text, images and video, LFM3-Audio-2B for text and audio, and LFM3-Omni-2B for text, image, video and audio inputs. Liquid AI also lists two decision models, d1-2B and d1-omni-2B, which it describes as producing calibrated answers. Each model branches from LFM3-1.7B-Base, with additional encoders or decoders for particular modalities.
Liquid AI describes the base model as a 24-layer tri-hybrid architecture that combines short convolutions, sliding-window attention and global attention. The company says the base was trained on 22 trillion tokens, including a final 3.5-trillion-token stage using curated data with more agent traces and tool calls. The text model supports a stated 256,000-token context window; the company also says it fits in about 1 GB at 4-bit precision.
The company reports that LFM3-1.7B outperforms its LFM2.5-2.6B predecessor while using 35% fewer parameters, and is about 25% faster than LFM2-1.7B at an 8,000-token context. Those comparisons are attributed to Liquid AI; the supplied material does not provide independent evaluations or full testing protocols. It says the models are available through Transformers, with model cards offering implementation and inference details.
Why Smaller Multimodal Models Matter
LFM3’s pitch is that developers could select a model suited to their input type and adapt it for a particular job, rather than relying on one general-purpose system for every task. If the company’s size, speed and quality claims hold in practical testing, local deployment could be useful for applications that need low latency, operate with limited computing resources or process data close to where it is generated.
The range of inputs also points to potential uses in interactive assistants, robotics and systems that need to interpret audio or video. Those are possible applications, not demonstrated outcomes in the announcement. The material does not establish real-world cost savings, privacy benefits or how the models perform after fine-tuning on a specific task. Actual results will depend on hardware, workload and the quality of the adaptation.
Products Worth Considering
Logitech MX Master 4 for Mac, Ergonomic Bluetooth Mouse with Haptic Feedback, Ultra-Fast Scrolling, USB-C Charging, for MacOS, iPadOS, Windo
DJI Osmo Mobile 8 3-Axis Smartphone Gimbal Stabilizer
Sony WH-1000XM6 Wireless Noise-Canceling Over-Ear Headphones, Blue, Bundle
As an affiliate, we earn on qualifying purchases.
Liquid AI presents LFM3 as the next generation of its Liquid Foundation Models, following the LFM2 family. The company says LFM2 paired short convolutions with attention to support efficient device use. For LFM3, it describes a hardware-in-the-loop architecture search across 14 combinations of devices and computing backends, including phones, a Raspberry Pi 5, laptops, cloud CPUs and GPUs, and Qualcomm’s Hexagon NPU.
The new base changes several design details, according to the company. Its convolution layers use a kernel size of seven, compared with three in LFM2, and it adds head-wise gates that can scale individual attention heads. Liquid AI also describes partial rotary position embeddings in global attention layers and normalization at each block’s input and output. These design choices are presented as ways to improve capability per parameter and support longer-context inference; their practical effect has not been independently established in the supplied material.
“LFM3 is our new generation of Liquid Foundation Models (LFMs), built to be the starting point for specialized intelligence.”
— Liquid AI
Benchmark Evidence Still Limited
The performance and efficiency figures in the supplied material come from Liquid AI. It does not include independent test results, complete evaluation protocols or enough detail to compare results across hardware and deployment conditions. The material begins to discuss vision benchmark results but does not provide the full comparison. As a result, the reported figures should not be treated as independent confirmation of performance.
Several deployment questions also remain open. The announcement does not specify how much fine-tuning data, time or memory a particular application will require, or how quality and latency change under production workloads. A stated context window and model size at 4-bit precision do not by themselves establish end-to-end speed, accuracy or operating cost. The supplied information also does not state release dates for every model, licensing terms or whether all model weights and deployment options are available.
Model Cards and Independent Testing
Developers can consult the LFM3 model cards and the agent-harness guide cited by Liquid AI for setup instructions, supported inference engines and implementation details. The supplied material says the models run in Transformers and includes an example for loading the text model.
The next useful evidence would be full benchmark methods and results, independent comparisons on the same hardware, and measurements of memory use, latency and output quality after task-specific fine-tuning. Those details would help determine whether the company’s reported efficiency translates into practical advantages for particular applications.
Key Questions
What is LFM3?
LFM3 is Liquid AI’s family of models built from a shared 1.7-billion-parameter base, with versions adapted for text, vision, audio, multimodal inputs and decision tasks.
Which inputs do the LFM3 models support?
The listed models cover text, images, video and audio in different combinations. LFM3-VL-2B handles text, images and video; LFM3-Audio-2B handles text and audio; and LFM3-Omni-2B combines all four input types.
Can LFM3 be fine-tuned and run on edge devices?
Liquid AI says the models are designed for fine-tuning on one GPU and deployment on edge devices or other infrastructure. The supplied information does not establish the hardware, memory or fine-tuning requirements for every application.
Have LFM3’s performance claims been independently verified?
Not in the supplied material. The reported speed, accuracy and efficiency figures are claims from Liquid AI; independent evaluations and complete benchmark protocols are not included.
Primary source: Hugging Face · via ThorstenMeyerAI.com
