模型发布/更新
-
OpenAI launches real-time translation model with support for 70+ input languages
OpenAI real-time translation feature — speak in more than 70 input languages, translated into 13 output languages:…
-
Xiaomi open-sources controllable video Foley sound generation model ControlFoley, letting sound follow your intent
Xiaomi's large model application team released ControlFoley, an open-source controllable video Foley sound generation…
-
Qwen-VLA: from understanding the world to taking action
Qwen has released Qwen-VLA, a general vision-language-action model built on the Qwen multimodal backbone that extends…
产品发布/更新
-
Codex can autonomously manage conversation threads and parallel tasks
Codex for managing the Codex interface: [quoting @guinnesschen]: If you are tired of managing Codex conversation…
-
Gemini Omni can bring sketches to life
Gemini Omni can even turn simple sketches into new realities. Try it yourself in the Gemini app. Upload a video of…
-
Codex now supports computer use on Windows
Windows users, this one is for you. Computer use is now available on Windows, so Codex can take actions on your…
-
Guardrails: protecting your agents, data and costs
Guardrails is a set of configurable safety and governance tools offering budget enforcement, zero data retention,…
-
ComfyUI now supports direct calls to OpenRouter models
You can now use your OpenRouter models directly in ComfyUI workflows! [quoting @ComfyUI]: ComfyUI just added…
-
OpenRouter supports model-generated file patches
OpenRouter now supports "apply_patch", a server tool that lets any model propose file edits via V4A diffs using the…
-
ChatGPT conversation table of contents now available
For every ChatGPT conversation that started as "just one quick question" and turned into a full-length saga: the table…
行业动态
-
Anthropic估值9650亿美元超越OpenAI
Anthropic在最新融资轮中筹集了650亿美元,融资后公司估值达到9650亿美元,首次超越竞争对手OpenAI的估值水平。
-
Four departments including CAC: improve AI literacy of the whole people, accelerate talent cultivation and deepen popularization
The Cyberspace Administration of China and three other departments jointly issued the 2026 Key Work Points for…
-
波士顿儿童医院利用AI解锁新诊断
波士顿儿童医院通过部署OpenAI技术,用于改善患者护理并减轻运营负担,成功帮助诊断了超过40种罕见病病例。
论文研究
-
GPIC: large-scale visual generation benchmark dataset released
I'm very excited about this visual generation benchmark dataset for the new era of large-scale generative models! 🤩
技巧与观点
-
Seeing is believing: incredible inference speed
The Kog team achieved extremely high single-user inference speeds on standard data center GPUs, reaching 3,000…
-
What happens when a company gets too "AI-addicted"?
Box founder Aaron Levie noted that the people deciding to replace employees with AI are often those least familiar…
-
Tesla's FSD safety claims questioned
Tesla claims its Full Self-Driving (FSD) software is up to 10 times safer than humans, but a Reuters investigation…
-
Cognition的Scott Wu表示:AI编程智能体不应取代人类
Cognition公司开发了Devvin,这是一个号称首个且最成功的AI编程智能体。其著名程序员创始人Scott Wu明确表示,该智能体并非旨在取代人类程序员。