Qwen3.8-Flash-Next-Base 6B参数模型领先8项基准测试

精选理由

阿里巴巴Qwen3.8-Flash-Next-Base模型参数少，性能强，是MMLU-Pro等基准测试的领先者，值得一看。

AI 摘要

Qwen3.8-Flash-Next-Base参数量仅为6B，在14项基准测试中领先8项，包括MMLU-Pro、SuperGPQA等，与Qwen3.7-Plus-Base在剩余测试中保持竞争力。其51B N-gram嵌入参数使用确定性查找，不增加每词矩阵乘法成本。

原文 · 阿里通义 Qwen

With only 6B active parameters, Qwen3.8-Flash-Next-Base tops 8 of 14 benchmarks, including MMLU-Pro,...

With only 6B active parameters, Qwen3.8-Flash-Next-Base tops 8 of 14 benchmarks, including MMLU-Pro, SuperGPQA, BBH and GSM8K. And it remains competitive with Qwen3.7-Plus-Base on the rest. Its 51B N-gram embedding parameters use deterministic lookups, adding no per-token matrix-multiplication cost. 💬 0 🔄 3 ❤️ 18 👀 1579 📊 3 ⚡

查看原推