阿里发布Qwen3.8-Flash：Qwen4架构预览，开放权重
⚡Meet Qwen3.8-Flash, a multimodal MoE and an early preview of the Qwen4 architec…
做模型选型和推理优化的同学必看，Qwen3.8-Flash用1/9成本超越前代，还提前曝光了Qwen4架构，赶紧去跑一下benchmark验证效果。
⚡Meet Qwen3.8-Flash, a multimodal MoE and an early preview of the Qwen4 architecture, now open-weight!
⚡ 认识 Qwen3.8-Flash，一款多模态 MoE 模型，也是 Qwen4 架构的早期预览版，现已开放权重！
The production version Qwen3.8-Flash will be available soon via QwenCloud API at just $ 0.16/1M input tokens and $ 0.47/1M output tokens.
生产版 Qwen3.8-Flash 即将通过 QwenCloud API 提供，价格仅为每百万输入 tokens $0.16，每百万输出 tokens $0.47。
125B parameters + 51B N-gram embeddings, with just 6B activated per token. Unmatched cost-efficiency.
125B 参数 + 51B N-gram 嵌入，每个 token 仅激活 6B 参数。无与伦比的成本效益。
What's new: 🥳 - Next architecture: GDN + QSA hybrid attention, Gated Residual, N-gram Embedding & Muon optimizer, serving as a precursor to the architecture used in Qwen4. - Dramatically lower training and inference costs: trained at just 1/9 the cost of Qwen3.7-Plus, while outperforming it across the board with especially strong gains in coding and office tasks. - Strong performance: scoring 58.7 on DeepSWE 1.1, 62.5 on SWE-bench Pro, 73.9 on CoWorkBench, 84.5 on AndroidWorld, and 95.7 on MathVision (with CI). - 262K native context, extensible to 1M with YaRN.
新特性：🥳 - 下一代架构：GDN + QSA 混合注意力、门控残差、N-gram 嵌入和 Muon 优化器，作为 Qwen4 所用架构的前驱。 - 大幅降低训练和推理成本：训练成本仅为 Qwen3.7-Plus 的 1/9，同时全面超越其性能，尤其在编码和办公任务上表现突出。 - 强劲性能：在 DeepSWE 1.1 上得分 58.7，SWE-bench Pro 上得分 62.5，CoWorkBench 上得分 73.9，AndroidWorld 上得分 84.5，MathVision（含 CI）上得分 95.7。 - 原生上下文长度 262K，可通过 YaRN 扩展至 1M。
We’re also releasing the weights for Qwen3.8-Flash-Next, giving the community an early look at the new architecture we’re exploring for Qwen4.🚀
我们还发布了 Qwen3.8-Flash-Next 的权重，让社区提前一窥我们为 Qwen4 探索的新架构。🚀
We can't wait to see what you build with Qwen3.8-Flash!👀👇 -Model Studio: https://click.alibabacloud.com/m/20000002820/ -QwenCloud: https://click.qwencloud.com/m/20000002828/
我们迫不及待想看到你用 Qwen3.8-Flash 构建的作品！👀👇 - Model Studio：https://click.alibabacloud.com/m/20000002820/ - QwenCloud：https://click.qwencloud.com/m/20000002828/
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力