精选60Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)模型发布/更新多源精选 ×8
GLM 5.3 Flash 18B 激活参数 MoE
GLM 5.3 Flash is a 18B active MoE, pretrained on 30T tokens
GLM 5.3 Flash is a 18B active MoE, pretrained on 30T tokens I think it was trained on Nvidia. but this is not, strictly speaking, necessary OpenPangu 2.0 pro is 505B 18B active, Ascend MooreThreads has some 236B MoE (12-22 AB?), 25T LongCat 2.0 – 1600B 49B 35T, Ascend this is no longer a barrier
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力