Liquid AI 发布开源模型 LFM2.5-24B-A2B
A small model drop from Liquid AI 👀
Liquid AI 发布开源权重模型 LFM2.5-24B-A2B，总参数 24B、每 token 激活 2B，上下文扩展至 131K tokens，词表 128K，预训练 33T tokens，并称其为推理模型。官方表示该模型在智能体与工具调用任务上可与总参数和激活参数均多约 50% 的 Qwen3.6-35B-A3B 竞争，面向笔记本和 PC 本地运行，基础与后训练权重已在 Hugging Face 开放。
推荐理由
AI 评分 55
Liquid AI 新模型以 2B 激活参数对标更大规模模型，读者可据此了解端侧智能体的参数效率取舍。
主题
查看原帖Today we release 𝙻𝙵𝙼𝟸.𝟻-𝟸𝟺𝙱-𝙰𝟸𝙱, the largest 𝙻𝙵𝙼𝟸.𝟻 family member. It is an efficient local model for laptops and PCs to work through multi-step tasks inside popular agent harnesses. > 24B total parameters, 2B active per token > Context expanded to 131K tokens > A 128K-token vocabulary > 33T pretraining tokens > It is a reasoning model >…
展开完整原文
Today we release 𝙻𝙵𝙼𝟸.𝟻-𝟸𝟺𝙱-𝙰𝟸𝙱, the largest 𝙻𝙵𝙼𝟸.𝟻 family member. It is an efficient local model for laptops and PCs to work through multi-step tasks inside popular agent harnesses. > 24B total parameters, 2B active per token > Context expanded to 131K tokens > A 128K-token vocabulary > 33T pretraining tokens > It is a reasoning model > Very easy to fine-tune and customize > Performs competitively with Qwen3.6-35B-A3B, a model with ~50% more total and active parameters, on agentic and tool-use tasks. > Open base and post-trained weights on @huggingface 🤗 1/
正文
当前显示原文，中文全文译文尚未就绪。
A small model drop from Liquid AI 👀
Liquid AI has released a new open-weight LFM2.5-24B-A2B model with 24B total parameters and 2B active parameters.
Based on benchmarks, LFM2.5-24B-A2B performs on par with the larger Qwen3.6-35B-A3B model.
The model is available on HF 🔥 https://t.co/ssrKPLq3hV