{"version":1,"type":"story","url":"https://digestai.news/story/lfm2-5-24b-a2b-built-for-fast-on-device-agents","json":"https://digestai.news/story/lfm2-5-24b-a2b-built-for-fast-on-device-agents.json","markdown":"https://digestai.news/story/lfm2-5-24b-a2b-built-for-fast-on-device-agents.md","slug":"lfm2-5-24b-a2b-built-for-fast-on-device-agents","headline":"LFM2.5-24B-A2B: Built for fast on-device agents","summary":"It is designed for fast, private agents that run entirely on device, with a reasoning‑only architecture that outputs a chain of thought before the final answer.\n\nIt scored 48.09 on LiveClaw versus 38.19 for the competitor, and 27.69 on AutomationBench versus 8.72.\n\nBoth the base and post‑trained versions are open‑weight and available today on Hugging Face and Liquid Console. The model supports llama.cpp, MLX, vLLM, SGLang, and ONNX, enabling edge inference on Apple, AMD, Qualcomm, and Nvidia hardware.","keyPoints":["It outperforms Qwen3.6‑35B‑A3B on 5 of 8 agentic benchmarks, scoring 48.09 on LiveClaw versus 38.19."],"whyItMatters":"The release shows that large‑parameter models can be engineered for on‑device inference, reducing latency and privacy concerns for mobile and edge AI. Its competitive performance on agentic benchmarks suggests that reasoning‑only architectures can match larger MoEs while using fewer active parameters, potentially lowering deployment costs for developers.","category":{"slug":"models","name":"Generative AI & Models","url":"https://digestai.news/category/models"},"entities":{"companies":["Liquid AI","Hugging Face"],"models":["LFM2.5-24B-A2B","Qwen3.6-35B-A3B"],"people":[]},"firstPublishedAt":"2026-10-14T23:59:37Z","updatedAt":"2026-10-14T23:59:37Z","sourceCount":1,"hasPrimarySource":false,"sources":[{"outlet":"liquid.ai","title":"LFM2.5-24B-A2B: Built for fast on-device agents","url":"https://liquid.ai/blog/lfm2-5-24b-a2b","publishedAt":"2026-10-14T23:59:37Z","type":"press","primary":false,"lead":true}],"sourceNotes":null,"discussions":[{"site":"Hacker News","url":"https://news.ycombinator.com/item?id=50097369","points":49}],"thread":{"title":"Open-Source AI Model Evolution Accelerates","url":"https://digestai.news/thread/openarch-launches-pytorch-repo-for-readable-llm-architecture-implementations","storyCount":7},"cite":{"text":"Digest AI, \"LFM2.5-24B-A2B: Built for fast on-device agents\", 14 October 2026, https://digestai.news/story/lfm2-5-24b-a2b-built-for-fast-on-device-agents","publisher":"Digest AI","title":"LFM2.5-24B-A2B: Built for fast on-device agents","datePublished":"2026-10-14T23:59:37Z","url":"https://digestai.news/story/lfm2-5-24b-a2b-built-for-fast-on-device-agents"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}