双语科技快报 | 2026-08-12 | 01:40 速览
## 1. Sources: Nvidia is developing a Nemotron 4 model with 1T+ parameters, up from Nemotron 3 Ultra's 550B parameters but smaller than leading Chinese open models
来源: The Information
时间: 1:20 PM
消息来源：英伟达正在开发一款参数超过1万亿的Nemotron 4模型，相比Nemotron 3 Ultra的5500亿参数有所提升，但仍小于中国领先的开源模型。
## 2. Sundar Pichai says Gemini has hit 1B+ MAUs, the company's fastest-growing product ever and its 14th product to hit the 1B-user mark
来源: Sundar Pichai / @sundarpichai
时间: 1:12 PM
桑达尔·皮查伊表示，Gemini已突破10亿月活跃用户，成为公司有史以来增长最快的产品，同时也是第14个达到10亿用户里程碑的产品。
## 3. Brad Lightcap, who most recently led OpenAI's special projects division and formerly was its COO, says he is leaving the startup “to start something new”
来源: Stephanie Palazzolo / The Information
时间: 12:31 PM
最近负责OpenAI特别项目部门、此前担任首席运营官的布拉德·莱特卡普表示，他将离开这家初创公司“去开创一番新事业”。
## 4. Nvidia releases Nemotron 3.5 Lightning, an open 30B-parameter MoE model, and NeMo Switchyard, an open-source model routing library for AI agents
来源: Kyt Dotson / SiliconANGLE
时间: 9:20 AM
Nvidia发布Nemotron 3.5 Lightning，一款开放的300亿参数MoE模型，以及NeMo Switchyard，一个面向AI代理的开源模型路由库
## 5. IBM and Together AI sign a $240M, multiyear deal to build an AI inference cluster on IBM Cloud, using Nvidia's HGX B300 systems, to support open-source models
来源: Anhata Rooprai / Reuters
时间: 8:40 AM
IBM与Together AI签署了一项价值2.4亿美元、为期多年的协议，在IBM云上构建AI推理集群，采用Nvidia的HGX B300系统，以支持开源模型。