美团 LongCat 2.0 发布,未附基准测试成绩
精选理由
美团发了 LongCat 2.0,但这次没给跑分,之前那版号称昇腾 910B 训练、对标 SoTA,实测却不咋地,这篇吐槽可以看看。
LongCat 2.0 由美团训练,上一代 LongCat-Flash 曾是在华为昇腾 910B 芯片上完成训练的大型模型之一,基准成绩接近当时 SoTA。此番 LongCat 2.0 发布未附带基准测试报告。发帖人称此前实测中该模型表现不佳,并推测若本次报告成绩,理论上应超过 Sol 等对比对象。
原文 · Teortaxes
LongCat 2.0 was the first truly large model trained on Ascends (910B to boot!), and boasted of benchmarks on par with contemporary SoTA. In the little tests I gave it, it was completely hopeless. I see they don't report benchmarks this time. Logically, would've been >Sol lmao https://t.co/gEBBZVkHnR