快讯
Lewis Tunstall 评 Beam 发布：中国模型出色但 token 效率低，Beam 推理效率领先
作者转发 Lewis Tunstall 对 Beam 发布的评论，称其附图是发布中最重要的内容，并认为中国模型表现很好但 token 效率糟糕（作者建议用最大推理预算跑一次评测体会）。随附 Reflection 部分表示，Beam 在其模型类别中推理效率领先：比 GLM 5.2 效率高 3-4 倍，比领先的西方开源模型高 4 倍以上，意味着 Beam 能更快、更低成本地完成任务。以上效率和倍数均为原文/Reflection 部分的表述，未经独立核实。
事件来源
查看原文
RT Lewis Tunstall @ COLM 🌉 This is the most important plot from the Beam release IMO. The Chinese models are great, but horribly token inefficient (try running an eval with max reasoning to feel the pain). I'm looking forward to a future where open models start competing on this axis! Reflection: Beam leads in inference efficiency for its model class -- 3-4x more efficient than GLM 5.2 and over 4x more efficient than leading Western open models. This means Beam completes tasks faster and cheaper.