阿里巴巴集团已开源其 Qwen3.8-27B 模型，该模型迅速获得普及，两天内在 Hugging Face 上下载量超过一百万次。该模型拥有不到 300 亿个参数，因其性能受到开发者的称赞，一些人因其超越 Claude Opus 4.6 Max 等模型在编码和代理基准测试中的能力而将其昵称为“本地 Opus 4.6”。值得注意的是，Qwen3.8-27B 可以在仅需 16GB VRAM 的消费级硬件上运行，使其能够轻松实现本地部署。 AI
影响 加速了本地部署和开发中功能强大的开源模型的采用。
排序理由 前沿实验室模型发布，附带系统卡
在 Mastodon — fosstodon.org 阅读 →
AI 生成摘要 · Google Gemini · 来自 6 个来源。 我们如何撰写摘要 →
Alibaba's Qwen team open-sourced Qwen3.8-27B, which topped Hugging Face's global trending chart within two days and passed one million downloads. Overseas developers nicknamed it the local Opus 4.6: under 30B parameters, it outperforms every model released four months ago includi…
Alibaba's new open-source Qwen3.8-27B model isn't just a tech upgrade—it's a geopolitical hedge. By packing frontier-class coding and agent power into a model that runs locally on just 17GB of RAM, Chinese developers are building a global ecosystem of edge-AI apps that are struct…
<h1> Qwen 3 Explained: Alibaba's Flagship Model — Performance, Pricing, and How to Access It </h1> <p>Alibaba's Qwen family has been the quiet powerhouse of the Chinese AI ecosystem for years. With Qwen 3, the flagship generation of 2026, it's no longer quiet — it's the model dev…
Alibaba has open-sourced Qwen3.8-27B, a model under 30B parameters that outperforms Opus 4.6 on coding and agent benchmarks. The model reached over one million downloads on Hugging Face within two days. https:// pandaily.com/qwen-3-8-27b-open -source-local-opus-4-6-apache-2-0-17g…
Alibaba’s new Qwen3.8-27B model isn't just open-source under Apache 2.0; it runs on a single consumer GPU with 16GB of VRAM and beats Claude Opus 4.6 Max in specific coding benchmarks. By giving away a highly optimized, edge-capable model for free commercial use, Alibaba is aimin…
Alibaba hat Qwen3.8 27B veröffentlicht, ein vision-fähiges Modell unter Apache-2.0-Lizenz, das sich laut Simon Willison gut lokal auf einem Rechner mit genug RAM betreiben lässt. Problem in der Praxis: Das Modell neigt standardmäßig zu exzessivem Overthinking, was Latenz und Toke…