Liquid AI releases LFM2-24B-A2B, a 24B parameter mixture-of-experts model
Liquid AI has released LFM2-24B-A2B, a 24-billion parameter mixture-of-experts model designed for text generation and conversational tasks. The model supports nine languages including English, Arabic, Chinese, French, German, Japanese, Korean, Spanish, and Portuguese.
Liquid AI has released LFM2-24B-A2B, a 24-billion parameter mixture-of-experts (MoE) model for text generation and conversational applications. The model is now available on Hugging Face.
Model Specifications
LFM2-24B-A2B is a text-generation model built on the LFM2 architecture. The model employs a mixture-of-experts approach, a technique that allows different parameter subsets to activate for different inputs, potentially reducing computational overhead compared to dense models of similar capacity.
The model supports nine languages: English, Arabic, Chinese, French, German, Japanese, Korean, Spanish, and Portuguese. This multilingual capability positions it for deployment across diverse linguistic regions.
Deployment and Access
The model is compatible with Hugging Face endpoints and is available for US-region deployment. Liquid AI has published the model under a custom license ("license:other"), meaning usage terms differ from standard open-source licenses—users should review the specific terms before deployment.
The model carries the tag "edge," suggesting optimization for edge deployment scenarios where computational resources are constrained.
Technical Context
Liquid AI's LFM2 architecture appears designed to balance model capacity with inference efficiency through its mixture-of-experts design. The 24-billion parameter scale represents a middle ground in current model sizing, larger than smaller instruction-tuned models but significantly smaller than frontier 70B+ parameter models.
The model includes references to arxiv:2511.23404, indicating academic documentation may be available for the underlying architecture.
What This Means
Liquid AI's release of LFM2-24B-A2B adds another option to the growing ecosystem of open and semi-open models. The mixture-of-experts approach and explicit edge-deployment tagging suggest targeting use cases where inference cost and latency matter more than maximum capability. The multilingual support and US endpoint compatibility indicate focus on commercial deployment rather than pure research release. Builders evaluating cost-efficient alternatives to larger models should benchmark this model against comparable MoE designs from other providers.
Related Articles
Thinking Machines Lab launches Inkling 2 Small, a 594B-parameter sparse MoE model, on Tinker
Thinking Machines Lab introduced Inkling 2 Small on November 5, 2026. The sparse Mixture-of-Experts model has 594 billion total parameters and 17 billion active parameters, and is available through Tinker with closed weights at launch.
Baidu open-sources ERNIE 5.5 Flash Base, an MoE model with 1M-token context for fine-tuning and inference
Baidu open-sourced ERNIE 5.5 Flash Base on November 5. It is a Mixture-of-Experts base model pretrained on 33.5 trillion tokens, with 128K pretraining context and up to 1M tokens in fine-tuning and inference. Baidu claims it sets state-of-the-art results among models of the same size.
Salesforce launches Koa Decisions, a Gemma-based model that classifies text, image and audio requests
Salesforce introduced Koa Decisions on November 6, 2026. The decision model is built from Google's open-weight Gemma and accepts text, images and audio. It is the next version of HyperClassifier.
Underdog releases Woof, a 31M-parameter open text-to-speech model that runs in a browser
Underdog's new text-to-speech model Woof has 31 million parameters and weighs 31.7 MB. The company says it runs on most devices without an internet connection, and the weights are open to download. Underdog demonstrated it with a browser game posted on November 5, 2026.
Comments
Loading...