余温 (@gkxspace) “我真是不敢相信！！！ Claude 4.6 Opus 曾是我的梦中情模啊，呜呜呜，现在居然被 27B” — TopicDigg
余温
@gkxspace
加入 January 2025
0 正在关注 0 粉丝
我真是不敢相信！！！
Claude 4.6 Opus 曾是我的梦中情模啊，呜呜呜，现在居然被 27B 的开源模型超越了。
这玩意儿可以直接跑在消费级显卡。
量化后只要 17-19GB，3090、4090 显卡都能跑，单张 5090 甚至能跑到 206.1 tok/s。
我觉得这种小模型的价值，真的可能比旗舰大模型还要大。
1、代码库、公司文档、客户资料、会议记录，全都可以留在本地，消费级显卡就能部署啊，兄弟们。
2、拿它 coding、读资料、整理文档、跑长期 Agent，从此没有 token 焦虑。
更牛逼的是，它才 27B。
按照这个速度再发展一年，我真觉得今天 Fable5 这一档的效果，很可能就在一张消费级显卡里跑起来了。。。
显示更多
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 Face：
- ModelScope：
显示更多
