腾讯混元开源 Hy4 preview:770B MoE、1M 上下文
🤝
腾讯 HY 开源 Hy4 preview:旗舰级 MoE,1M 上下文,是其迄今最大的一次代际跃升。Apache 2.0。🔥 🤖 https://t.co/ZOc55cqdAC ⚙️ 总参数 770B,每 token 激活 49B。原生内置 MTP 层,开箱即支持投机解码。 📊 163 位内部专家对 203 个真实工程任务进行盲评:略胜 GLM 5.3(平均 2.99 对 2.92)和 Kimi K3(2.99 对 2.94)。 💼 基于真实生产工作流训练:软件工程、办公与数据分析、游戏开发以及科学研究。 🌍 面向全球用户:https://t.co/qQG4sOFpfV
查看原帖Tencent HY open-sourced Hy4 preview: a flagship MoE with 1M context, and their biggest generational leap measured to date. Apache 2.0.🔥 🤖 https://t.co/ZOc55cqdAC ⚙️ 770B total params, 49B active per token. Ships with a native MTP layer for speculative decoding out of the box. 📊 Blind eval by 163 internal experts across 203 real engineering tasks: it…
展开完整原文
Tencent HY open-sourced Hy4 preview: a flagship MoE with 1M context, and their biggest generational leap measured to date. Apache 2.0.🔥 🤖 https://t.co/ZOc55cqdAC ⚙️ 770B total params, 49B active per token. Ships with a native MTP layer for speculative decoding out of the box. 📊 Blind eval by 163 internal experts across 203 real engineering tasks: it edges out GLM 5.3 (avg 2.99 vs 2.92) and Kimi K3 (2.99 vs 2.94). 💼 Trained on real production workflows: software engineering, office and data analysis, game development, and scientific research. 🌍 For global users: https://t.co/qQG4sOFpfV
正文
🤝