30 Oct 2026
Fake coding extension in Cursor's marketplace planted an SSH key
A tampered Anthropic.claude-code 2.1.313 sat in the Cursor extension marketplace for about 11 hours on 29 October. Cursor says at most…
Liquid AI's open-weight 1.7-billion-parameter text model, released 27 October 2026 as the base of the LFM3 family, with a 256K context, about 1 GB at 4-bit, and fine-tuning on a single GPU.
Not yet independently verified. Rests on Liquid AI's own announcement of 27 October 2026 and its Hugging Face model card. The comparison with LFM2.5 is Liquid AI's own claim, no independent dated test was found, and we have not summarised the licence terms beyond its name. We will update this when it can be confirmed, and remove this note.
LFM3-1.7B is the text model of LFM3, the third generation of Liquid Foundation Models that Liquid AI announced on 27 October 2026. The first LFM3 release covers six models in the 2-billion-parameter class, for text, vision, audio, omni and decision tasks, all trained from one shared base, LFM3-1.7B-Base. Base and post-trained weights are open on Hugging Face, and the models are also offered in Liquid's Playground and through the Liquid AI API.
Liquid AI builds small models meant to run on the device where an application lives. LFM3 uses a new tri-hybrid architecture, redesigned with the company's hardware-in-the-loop architecture search for agentic use and fast inference over long contexts, and was pre-trained on 22 trillion tokens. Liquid AI says LFM3-1.7B outperforms its previous LFM2.5-2.6B while using 35 percent fewer parameters.
The text model has a 256K-token context window and fits in about 1 GB at 4-bit precision. Its Hugging Face card lists English, Arabic, Chinese, French, German, Italian, Japanese, Korean, Portuguese, Spanish, Vietnamese, Thai, Indonesian, Hindi, Russian and Polish.
The model is meant as a starting point for specialised models: Liquid AI says every LFM3 model can be fine-tuned on a single GPU and deployed where the application runs, which suits teams building on-device assistants, offline tools or narrow agents that cannot send data to a hosted model. The weights are released under the LFM Open License v1.0, which has its own commercial-use terms that a business should read before deployment.
Maker: Liquid AI Released: 27 October 2026 Parameters: 1.7 billion Context: 256K tokens Size: about 1 GB at 4-bit precision Licence: LFM Open License v1.0, open weights Access: Hugging Face, Liquid Playground, Liquid AI API
| Benchmark | Official | Community avg |
|---|---|---|
| No benchmark scores yet. Be the first to add one. | ||
“Official” values are editor-approved and feed the ranking. “Community avg” is the mean of member submissions (shown for transparency; it never affects the ranking until an editor approves a value).
Sign in to add a benchmark score for this model.
Used it? Your experience helps other buyers decide.
Write a reviewNo questions yet. Be the first to ask about LFM3-1.7B.
Everything here links back to the same verified catalogue. Pick your next stop.