True bidirectional real-time conversation is here! OpenAI has quietly begun testing its new voice model, 'Bidi 1', on the ChatGPT web and mobile apps. Featuring full-duplex communication, Bidi 1 represents the most significant upgrade to ChatGPT's voice capabilities since its inception. 🔹 **Full-Duplex Interaction*
真正的双向对谈来了！ChatGPT 网页与 App 端已流出全新语音模型“Bidi 1”的测试入口。 该模型支持全双工通信，是 ChatGPT 语音上线以来的最大升级。 🔹 **全双工交互**：支持随时插话打断。Bidi 1 允许 AI 在说话的同时保持监听。 用户在中途发出指令，模型会立即中断当前发言并响应新命令。 🔹 **界面调整**：新增黄色对话气泡。Bidi 1 已出现在部分用户的模型列表中，与标准和高级语音并列，选中后气泡转为黄色。 🔹 **自然沟通**：消除人机对话等待。模型体验向真实人类交流习惯看齐，实现了无延迟、更连贯的无缝口语互动。 原文链接: https://www.ithome.com/0/967/852.htm
Priced at just $299! Meta has launched its first line of self-branded AI smart glasses, which are $80 cheaper than the latest Ray-Ban Meta models starting at $379. This product launch represents a crucial step for Meta to drive consumer adoption of wearable AI technology and demonstrate concrete
售价仅 299 美元！Meta 推出首款自研 AI 眼镜，比雷朋联名款便宜 80 美元。 这是 Meta 抢占可穿戴设备大众市场、回击对其 AI 巨额投入质疑的关键举措。 🔹 **自研镜框**：摆脱对雷朋的依赖。新眼镜由 Meta 主导设计，提供 Adventurer、Fury 及 Kylie 联名款三种风格，陆逊梯卡仅合作镜片。 🔹 **多模态升级**：搭载 Muse Spark AI 模型。支持听歌、翻译及基于镜头的环境问答，能提取照片细节并记忆用户偏好。 🔹 **商业策略**：提供更低价格段。Meta 通过更亲民的价格扩大用户基数，验证其大模型商业化的可行性。 原文链接: https://cnnespanol.cnn.com/2026/06/23/
Generating a 5-second 480P video end-to-end takes just 1.8 seconds on a single NVIDIA GeForce RTX 5090! UCSD's Hao AI Lab has officially open-sourced FastWan-QAD, a family of distilled video generation models that significantly outperform existing solutions like TurboDiffusion and LightX2V. Built
单张 5090 仅需 1.8 秒即可端到端生成 5 秒 480P 视频！UCSD 团队开源了 FastWan-QAD 视频生成模型，性能超越 TurboDiffusion。 项目基于 Wan2.1-T2V-1.3B，通过量化感知蒸馏（QAD）技术将推理速度推向极限。 🔹 **优化技术**：极致量化与高密度注意力。模型线性层被量化为 NVFP4 或 FP8 精度，搭配 SageAttention3 FP4 后端。 拒绝稀疏路由，保持 100% 密集注意力以保障视频质量。 🔹 **开源版本**：本次推出三款模型。包含适配 5090 的 NVFP4 旗舰版（1.8秒）
From physical hardware blockades to direct cloud software restrictions. The U.S. Department of Commerce's Bureau of Industry and Security (BIS) recently issued an emergency export control directive targeting AI startup Anthropic, legally banning all foreign nationals, both inside and outside the United States, from accessing its advanced