Ant Ling· @AntLingAGI · X·· 2026-05-17精选AI 评分76
AI 导读
蚂蚁集团发布万亿参数推理模型Ring-2.6-1T,专为现实世界智能体工作流构建。该模型采用MIT许可,上下文长度通过YaRN技术从128K扩展至256K,并采用Async RL与IcePop混合训练架构。其核心特点是具备双推理努力模式:“high”模式用于快速智能体循环,“xhigh”模式用于深度推理,旨在实现更好的成本与性能平衡。模型已开源,欢迎社区反馈。
推荐理由
蚂蚁把1T推理模型以MIT许可开源,还专门为Agent循环做了高低推理双模式,做开源Agent的团队直接多了一个强力基座可选。
正文 · 原文
Thanks @AdinaYakup and the @huggingface community for the continued recognition!
We feel happy to bring another 1T thinking model to the community! Comments and feedbacks welcome!
Ant group just dropped Ring-2.6-1T 🔥 1T reasoning model, built for real world agent workflows. ✨ MIT license ✨ 128K >> 256K context (YaRN) ✨ Async RL + IcePop training architecture ✨ Dual reasoning effort: "high" for fast agent loops, "xhigh" for deep reasoning = Better cost/performance tradeoff 👀在 X 查看被引用的帖子
来源:Ant Ling · x.com