d1-3B

d1-3B is the larger of two open-weight decision models that Liquid AI released on 7 October 2026 as the d1 family. It is built on the company’s LFM2.5-VL-3B vision-language model, has about 3.12 billion parameters and a context length of 32,768 tokens, and accepts text and images. It is not a chat model: it gives a single-pass, calibrated answer such as yes or no, a choice from a list or a score. Liquid AI lists routing and triage, moderation, intent and topic classification, extraction checks, reranking, LLM-as-a-judge scoring, agent guardrails and visual inspection as intended uses. The weights are on Hugging Face under the licence tagged lfm1.0 and can be downloaded, fine-tuned and deployed. Liquid AI reports a mean of 82.9 across seven public text benchmarks that it ran itself, and a Decision Index v0.2.1 public-split score of 48.57 without saying who measured it. It also published latency figures on devices from Apple, NVIDIA and AMD. Those are vendor figures, so no score is stored here. No price is stated.

Not yet independently verified: Read on Liquid AI’s own announcement dated 7 October 2026 and the Hugging Face model card; the only dated independent mention found is a Hacker News post on 8 October. The scores in the announcement are Liquid AI’s own, apart from a Decision Index figure whose operator the page does not name, so no score is stored. The model card gives no release date. We will update this when it can be confirmed, and remove this note.