阿里把 27B 开源了，原生多模态，整体超 Qwen3.7-Plus，17GB 内存就能本地跑，适合自己部署。
阿里开源了 Qwen3.8-27B，这是原生多模态稠密模型，27B 参数在综合表现上超过 Qwen3.7-Plus。该模型原生支持 262K 上下文，可通过 YaRN 扩展到 1M。它兼容 vLLM、SGLang、TokenSpeed，并提供官方 FP8 版本。Unsloth 的量化方案只需 17GB RAM 即可本地运行。此外，Max 级 Qwen3.8-2.4T-A95B 的权重也已开放。
Qwen3.8-27B 开源了 ⚡ 相比 Qwen3.8 Max，27B 开源更值得特别关注！ Qwen3.8-27B 是原生多模态稠密模型，表现整体超过 Qwen3.7-Plus，在真实编程与办...
Qwen3.8-27B 开源了 ⚡ 相比 Qwen3.8 Max，27B 开源更值得特别关注！ Qwen3.8-27B 是原生多模态稠密模型，表现整体超过 Qwen3.7-Plus，在真实编程与办公流程上更突出。 既然是 27B，就得好好看看部署环境要求了，兼容 vLLM / SGLang / TokenSpeed，另有官方 FP8。社区已有 GGUF / Unsloth 等量化。 这里推荐 @UnslothAI 的量化方案，17GB RAM 可跑： x.com/UnslothAI/stat… Qwen @Alibaba_Qwen We promised open weights for Qwen3.8. Now, time to meet them! 🎉 ⚡ Qwen3.8-27B: - A native multimodal dense model. With just 27B parameters, it outperforms Qwen3.7-Plus overall and shines in real-world coding & office workflows. - 262K native context, easily extendable to 1M tokens via YaRN. - Built for builders. Highly efficient, high-quality, and licensed under Apache 2.0. 🚀 The open weights for Qwen3.8-2.4T-A95B (Max-level) have also been released recently. Whether you're shipping lightweight applications with Qwen3.8-27B locally or building agents with Qwen3.8-2.4T-A95B, they're yours now! Download, deploy, and build something we haven't imagined yet. 👀👇 - Hugging Fa huggingface.co/collections/Qw… YEVj - ModelSco modelscope.cn/collections/Qw… GkhC 🔗 View Quoted Tweet 💬 0 🔄 0 ❤️ 0 👀 385 ⚡