Google DeepMind 试点首个双盲 AI 评测
DeepMind 借助密码学安全环境开展双盲评测，以提升闭源模型基准测试结果的可信度。
原文来源：Google DeepMind（deepmind.google）。本页摘要与要点整理自公开来源，版权归原作者所有，完整内容请阅读原文。
Gemini Omni 1.1 Flash lets you build with more controlGemini Omni 1.1 Flash 发布，为开发者提供更多视频创作控制
Google DeepMind · 2026-08-28
Google’s AI Mode can now track flight prices, help book hotels, and moreGoogle 搜索 AI Mode 新增机票价格追踪与酒店预订功能
TechCrunch · 2026-08-28
Anthropic's new hardware standard lets AI agents control the physical worldAnthropic 推出硬件标准，让 AI Agent 控制物理设备
Ars Technica · 2026-08-28
Anthropic continues compute-gobbling streak in $45B deal with NscaleAnthropic 与 Nscale 签下约 450 亿美元算力租赁协议
TechCrunch · 2026-08-27
Our decision on Cursor following its acquisition by SpaceXOpenAI 宣布将逐步终止向 Cursor 提供模型的合同
OpenAI · 2026-08-28
Intelligent transcription with Gemini 3.5 TranscribeGoogle 发布语音转文字模型 Gemini 3.5 Transcribe
Google DeepMind · 2026-08-27