AI 导读
Qwen 发布 Qwen3.8-Flash-Next，这是一个 125B 参数、含 51B n-gram 嵌入向量、每 token 仅激活 6B 参数的稀疏 MoE 开源模型。在 9 项可比基准中有 8 项超过 Claude Opus 4.6 Max，包括 62.5 SWE-bench Pro、81.0 SWE-bench Multilingual、73.9 CoworkBench、81.3 IFBench、91.7 GPQA Diamond 和 91.9 LiveCodeBench。该模型在表中多数项目上也优于 Qwen3.8-27B 与 DeepSeek-V4-Flash。
正文
official: https://t.co/mzQDqiK6Rq
Qwen 3.8 Flash-Next official released: A 6B-active open model just beat Claude Opus 4.6 Max across 8 of 9 comparable benchmarks! Qwen3.8-Flash-Next is a highly sparse MoE: • 125B model parameters • 51B additional n-gram embeddings • Only 6B parameters active per token It scores: • 62.5 SWE-bench Pro • 81.0 SWE-bench Multilingual • 73.9 CoworkBench • 55.7 JobBench • 73.5 Toolathlon • 81.3 IFBench • 91.7 GPQA Diamond • 91.9 LiveCodeBench It also outperforms Qwen3.8-27B and DeepSeek-V4-Flash across most of the table. Super cool release!!在 X 查看被引用的帖子
来源：@kimmonismus · x.com