Liquid AI releases two lean encoder models that read long documents fast, even on a laptop
LFM2.5-Encoder-230M and LFM2.5-Encoder-350M can scan an entire contract in under 30 seconds on ordinary computer hardware, no specialist chip required.
Key points
- Liquid AI released two encoder models, LFM2.5-Encoder-230M and LFM2.5-Encoder-350M, on Hugging Face on an unspecified date in 2026.
- The smaller 230M model runs roughly 3.7 times faster than the comparable ModernBERT-base model on a standard laptop CPU at long inputs.
- Both models handle up to 8,192 tokens, roughly six to eight pages of dense text, in a single pass.
- LFM2.5-Encoder-350M ranked fourth out of 14 models tested, beaten only by models up to ten times its size.
- Both models are open-weight and free to download from Hugging Face today.
Liquid AI, a Boston-based AI company, has released two new encoder models, a type of AI model that reads and understands text rather than generating it from scratch. The models are called LFM2.5-Encoder-230M and LFM2.5-Encoder-350M, and they are available now on Hugging Face, the open-source AI platform where researchers and developers share models.
If you have never heard of an encoder, think of it this way. A chatbot like ChatGPT writes new sentences word by word. An encoder does the opposite job: it reads a piece of text and decides what it means, whether that is labelling an email as spam, flagging personal data in a document, or routing a customer query to the right department. Businesses run these tasks constantly, all day long, which makes speed and cost the deciding factors.
Why does running on a laptop CPU matter?
Most cutting-edge AI models need expensive specialised chips called GPUs to run at useful speeds. These encoders do not. On an ordinary laptop processor, the 230M model processes a full 8,192-token document, roughly six to eight pages of text, in about 28 seconds. The closest competing model, ModernBERT-base, takes over 90 seconds on the same hardware for the same length.
That 3.7-times speed gap is the headline claim, and it matters most for smaller teams or companies that do not want to pay cloud GPU bills for routine classification work.
How good are the results?
Speed means little without accuracy. Liquid AI tested both models across 17 tasks drawn from GLUE and SuperGLUE, two standard benchmark suites that researchers use to score how well a model understands language. Out of 14 models in the comparison, LFM2.5-Encoder-350M placed fourth. Every model ahead of it was larger, including one nearly ten times its size at 3.5 billion parameters.
The smaller 230M model beat ModernBERT-base and every variant of EuroBERT, despite being physically smaller than most of them.
| Model | Parameters | GLUE/SuperGLUE rank |
|---|---|---|
| LFM2.5-Encoder-350M | 350M | 4th of 14 |
| LFM2.5-Encoder-230M | 230M | beats ModernBERT-base |
| ModernBERT-base | ~149M | below both LFM2.5 models |
What can someone actually build with this?
Liquid AI lists several ready-to-run demos: a tool that spots and removes 40 types of personal information across 16 languages, a policy checker that scans text against company rules written in plain English, and a prompt router that sends user requests to the right service automatically. All demos run in a browser with no setup.
Developers can also download either model and adapt it to a specific task, a process called fine-tuning, using Hugging Face's standard tools. Liquid AI published a tutorial walking through that process on long legal documents.
Common questions
Do I need a powerful computer to use these models?
No. Both models are designed to run on a standard laptop or server CPU. You do not need a GPU, though using one will make them faster still at shorter documents.
Are these models free to use?
Yes. Both LFM2.5-Encoder-230M and LFM2.5-Encoder-350M are open-weight, meaning anyone can download and use them at no cost, including for commercial projects, subject to the licence terms on Hugging Face.
Which size should I choose?
Liquid AI's own guidance is straightforward: pick the 350M model when accuracy is the priority, and the 230M model when you need to process more documents per hour or your hardware is limited.
