全国产算力跑通!零NVIDIA含量:美团万亿大模型跻身全球第一梯队

快科技
Jul 03

快科技7月3日消息,前不久美团正式发布新一代基础大模型LongCat-2.0,采用MoE架构,总参数1.6万亿,每个Token激活参数约480亿。

这款模型最特殊的地方不在于性能顶尖,而在于从训练到推理全流程由国产算力集群完成,英伟达含量为零,这不仅是国产芯片+国产模型真正意义上的跑通,还直接挤进了全球第一梯队。

正式发布前,LongCat-2.0曾匿名Owl Alpha接入OpenRouter平台,总调用量冲进前三,在Claude Code Agent场景下,其调用量排名全球第二,仅次于Claude Opus 4.8。

但客观来看,LongCat-2.0并非性能最强的模型,技术社区评估其Agent核心能力接近Claude Opus 4.6,落后于最新的Opus 4.8;Coding能力略优于智谱GLM-5.1,但落后于6月新发布的GLM-5.2。

然而加上“英伟达含量为零”这个限定词后,这些相对劣势便构成了质的突破,此前国内大模型不缺国产叙事,有用国产芯片做推理的,也有做后训练的,但训练加推理全流程跑通万亿规模尚属首次。

正如一位开发者所言,之前相当于房子建好用国产算力装修,这次是从打地基到入住全程国产算力,住着居然很好。

Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Most Discussed

  1. 1
     
     
     
     
  2. 2
     
     
     
     
  3. 3
     
     
     
     
  4. 4
     
     
     
     
  5. 5
     
     
     
     
  6. 6
     
     
     
     
  7. 7
     
     
     
     
  8. 8
     
     
     
     
  9. 9
     
     
     
     
  10. 10