喵~ 今日重大新闻来啦！澜舟科技的小尾巴摇出了大惊喜，他们家的孟子3-13B大模型喵，已经正式开源啦！这意味着学术研究的小鱼干可以尽情享用，而且商家们也可以免费使用哦，真是太棒了！这款模型可不是一般的轻巧，它拥有万亿token的数据训练，智慧满满呢。
孟子3-13B在MMLU、GSM8K、HUMAN-EVAL这些大考场上，都展示出它的聪明才智，成绩亮眼。尤其是在轻量级模型的领域里，它中文英文切换自如，就连复杂的数学和编程问题也能轻松应对，简直就像猫咪玩线球一样灵活。据量子位的消息，这款模型在20B参数量以内的较量中，可是佼佼者呢！
这下子，无论是学术界的猫头鹰还是商业界的狐狸，都可以借助孟子3-13B的大脑，探索更多的知识森林了。一起期待它带给我们更多智慧的火花吧！喵~
英语如下：
Headline: “LanZhou Tech Stuns with Open-Source Release: Mengzi 3-13B Large Model Leads AI into a New Era, Free for Commercial Use!”
Keywords: Mengzi 3-13B, Open-source large model, Outstanding performance
News Content: Meow~ Big news today, everyone! LanZhou Tech has shaken its tail with a fantastic surprise – their Mengzi 3-13B large model has officially gone open-source! Now, academic researchers can enjoy it like treats, and businesses can use it for free, too! It’s pawsome! This model isn’t just lightweight, it’s been trained with trillions of tokens, making it super smart.
Mengzi 3-13B has proven its intelligence on major benchmarks like MMLU, GSM8K, and HUMAN-EVAL, scoring impressively. In the realm of lightweight models, it’s bilingual, switching between Chinese and English like a pro, and even tackles complex math and coding problems with ease, just like a kitty playing with a yarn ball. According to Quantum Bit, it’s a top performer among models with 20B parameters or fewer!
Now, both scholarly owls and business-savvy foxes can harness Mengzi 3-13B’s brainpower to explore the vast knowledge forest. Let’s look forward to the sparks of wisdom it will bring! Meow~
【来源】https://mp.weixin.qq.com/s/L4wqVnbS8a9FT0Sd8JvDPw
Views: 1