{"apiVersion":"v1","taxonomyVersion":1,"generatedAt":"2026-10-29T14:03:07.821Z","data":{"recordId":"wsa_ex4n73dpi33f","score":0.852,"confidence":"high","capped":false,"capReason":null,"sources":[{"n":1,"url":"https://huggingface.co/LiquidAI/LFM2.5-VL-3B","publisher":"Liquid AI","authority":"owner-controlled","verification":"verified","lastRetrievedAt":"2026-10-29T00:46:41.159Z","detectedPublishedAt":"2026-08-11T22:54:10.000Z","rights":"original"},{"n":2,"url":"https://www.liquid.ai/blog/lfm2-5-vl-3b","publisher":"Liquid AI","authority":"owner-controlled","verification":"verified","lastRetrievedAt":"2026-10-29T00:46:41.159Z","detectedPublishedAt":"2026-08-12T00:00:00.000Z","rights":"original"}],"claims":[{"id":"cd82052c-c51c-499a-a3f8-cd3de880f829","kind":"availability","statement":"LFM2.5-VL-3B is available on Hugging Face and Liquid AI Playground.","citations":[{"source":2,"supportProbability":0.96,"supports":true,"quote":{"text":"The model is available today on Hugging Face and our Playground .","rights":"original"}}]},{"id":"c466af28-34ea-4360-809e-7ece1303f0d3","kind":"capability","statement":"The model improves screen understanding, grounding, function calling, and multi-image input over LFM2-VL-3B.","citations":[{"source":2,"supportProbability":0.97,"supports":true,"quote":{"text":"LFM2.5-VL-3B builds on our previous LFM2-VL-3B , with significant improvements in screen understanding, grounding, function calling, and multi-image input.","rights":"original"}}]},{"id":"0fc0c89a-0b68-4b75-85f4-dcbe7980e95e","kind":"capability","statement":"The model is non-reasoning and answers directly to keep latency low.","citations":[{"source":2,"supportProbability":0.99,"supports":true,"quote":{"text":"LFM2.5-VL-3B is a non-reasoning model that answers directly, keeping latency low for real-time and on-device applications.","rights":"original"}}]},{"id":"97f4629f-f16c-4f8c-8216-2ae7e8a0024d","kind":"privacy","statement":"The WebGPU demo runs entirely on-device with no setup required.","citations":[{"source":2,"supportProbability":0.98,"supports":true,"quote":{"text":"The model runs entirely on-device with no setup required.","rights":"original"}}]},{"id":"0baab99e-f22d-4b13-b725-751ea241e3ee","kind":"recommendation-relevant","statement":"The model is open-weight and can be downloaded, fine-tuned, and deployed without restrictions.","citations":[{"source":2,"supportProbability":0.93,"supports":true,"quote":null}]},{"id":"40648cd9-da8a-478a-8aed-a116e1b40f55","kind":"capability","statement":"The model processes both text and images.","citations":[{"source":1,"supportProbability":0.97,"supports":true,"quote":{"text":"LFM2.5-VL-3B can process both text and images, and uses the LFM2.5-2.6B language model as its backbone, combined with a SigLIP2 NaFlex vision encoder.","rights":"original"}}]},{"id":"84ceb235-63bf-4bc4-a5b4-efd75b11ffb9","kind":"capability","statement":"It supports object detection and grounding with natural-language queries.","citations":[{"source":1,"supportProbability":0.92,"supports":true,"quote":{"text":"* **Better grounding**: Improved grounding and object detection with natural language queries.","rights":"original"}},{"source":2,"supportProbability":0.71,"supports":true,"quote":null}]},{"id":"6ee0a815-5ce0-4001-be9c-5aa78d892a56","kind":"capability","statement":"It supports full-page OCR with layout annotation.","citations":[{"source":1,"supportProbability":0.96,"supports":true,"quote":{"text":"* **Better OCR**: Full page OCR with layout annotation.","rights":"original"}}]},{"id":"c66d7a7f-9061-4ea1-9f96-fa64a746805a","kind":"availability","statement":"The original checkpoint and GGUF, ONNX, and MLX exports are linked for download.","citations":[{"source":1,"supportProbability":0.45,"supports":false,"quote":null}]},{"id":"66976905-8819-41c7-add2-8849636acf83","kind":"deployment","statement":"The model family is designed for on-device deployment.","citations":[{"source":1,"supportProbability":0.62,"supports":false,"quote":{"text":"LFM2.5-VL-3B is a multimodal variant of LFM2.5, a family of hybrid models designed for **on-device deployment**.","rights":"original"}}]},{"id":"5fc1eff5-947b-46b9-b847-ee035819c9b5","kind":"recommendation-relevant","statement":"Liquid AI recommends it for single-turn, high-throughput, low-latency tasks and not for long-context, reasoning-intensive tasks.","citations":[{"source":1,"supportProbability":0.97,"supports":true,"quote":{"text":"We recommend using it for single-turn, high-throughput, low-latency tasks; for example, for near-realtime object detection in automotive applications, batch processing scanned documents with OCR with layout information for turning PDFs into searchable text, or for on-device…","rights":"original"}}]}]},"nextCursor":null,"license":{"catalog":{"name":"CC BY 4.0","spdx":"CC-BY-4.0","url":"https://creativecommons.org/licenses/by/4.0/","attribution":"wowsuchai","attributionUrl":"https://wowsuchai.withglint.app","covers":["catalog metadata","classifications","generated summaries","feed event metadata"]},"excluded":{"covers":["source quotations","linked source content"],"terms":"Not licensed by wowsuchai. Source quotations and linked source content keep their original rights."}}}