Liquid AI releases LFM3-VL-2B on-device vision-language model
Original title: LiquidAI/LFM3-VL-2B — new model trending #30 on Hugging Face
The 1.92B model takes text, images, and video, reasons step by step, and uses the LFM Open License v1.0.
Liquid AI published LFM3-VL-2B, a post-trained vision-language model for on-device visual understanding, now trending on Hugging Face. It pairs the LFM3 language model with an in-house vision encoder and accepts text, images, and video. Liquid AI says it is the first LFM vision model with video input, plus step-by-step thinking and spatial reasoning, and is not recommended for coding, multi-turn chat, or complex reasoning.
- Parameters total 1.92B: 1.73B language model, 164M vision encoder, and 25M projector.
- Context length is 128K tokens with a 128,000-token vocabulary.
- Video is sampled at 2 fps, with 4 to 768 frames by default.
- The card lists 16 languages, including English, Arabic, Chinese, and Hindi.
- License is LFM Open License v1.0.
- Vendors
- Liquid AI
- AI models
- LFM3-VL-2BLFM3
From the original article
9 min read at huggingface.coimage-text-to-text · 1,596 downloads · 145 likes · license other
By LiquidAI · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.