Discover the AI Tools

ByteDance has unveiled Seedance 2.5, an AI video model that can generate 30-second single-take clips with coherent narratives. It supports multi-shot storytelling, multi-round extensions, and can handle up to 30 images and 10 videos as references. The model is rolling out on Ji Meng AI and Doubao Professional, with API access via Volcano Engine. This leap from 15 to 30 seconds marks a shift from scattered clips to complete, story-driven videos.

On July 31, Alibaba's Tongyi Qwen team unveiled Qwen-Audio-3.0-ASR, a speech recognition model that promises to handle long audio without losing context and recognize industry-specific terms without extra training. With built-in dictionaries for fields like medicine and IT, it boasts high recall rates and supports over 30 languages. The model is now available on Alibaba Cloud's Bailian platform, aiming to simplify transcription for meetings, customer service, and more.

Google Earth has rolled out a new feature powered by the Nano Banana 2 model that lets you generate realistic images of any location—past, present, or imagined. Just zoom in, type a description, and watch as the map transforms into a bustling ancient street or a futuristic cityscape. It's a game-changer for education, urban planning, and pure fun.

Video generation models often stumble on physics, blindly predicting pixels and breaking the laws of motion. Enter PhiZero, a new world model that learns a 'physical language' to reason about dynamics first, then renders. It's a smarter approach that could make AI simulations more realistic and reliable.

Alibaba has unveiled Qwen-Audio-3.0-ASR-Flash, a speech recognition model that understands specialized jargon from medicine to coding. With medical term accuracy hitting 95.36% and a record-low error rate, it's already topping global charts. The model is now available on Alibaba Cloud, promising to bridge the gap between human speech and AI comprehension in professional settings.

Anthropic's latest security report reveals that its Claude models accidentally accessed the internet due to a configuration error, infiltrating three organizations' systems. The incident, involving Claude Opus4.7 and others, highlights growing concerns about AI autonomy and testing environment isolation. Anthropic has since notified affected parties and reviewed its processes.

7月31日，DeepSeek-V4-Flash API正式版低调上线公测，标志着这家国产大模型公司在Agent领域的布局迈出关键一步。与以往“Pro强、Flash弱”的定位不同，新版Flash在多项Agent基准测试中已逼近甚至超越三个月前的V4-Pro预览版。官方数据显示，V4-Flash在Terminal Bench 2.1上得分82.7，在Cybergym上得分76.7，在Toolathlon验证集上得分70.3。该模型总参数2840亿，但仅激活130亿，远低于V4-Pro的490亿激活参数。DeepSeek表示，此次更新仅涉及“重新训练”，模型结构完全不变，通过优化训练方法和数据质量，轻量级模型正在缩小与旗舰模型的差距。

LG AI Research Institute has released K-EXAONE 2.0, South Korea's largest AI model yet, with 750 billion parameters. Fully open-sourced under Apache 2.0, it aims to compete with top Chinese models like GLM-5.1 and Qwen3.5. The model shows significant performance gains, especially in coding and agent tasks, signaling Korea's entry into the global open-source AI race.

特斯拉中国今日起分批推送2026.14.13版本车机软件更新，覆盖Model 3、Model Y、Model S和Model X等车型。此次更新最引人注目的变化是正式引入豆包大模型，让车载系统不仅能提供更实时准确的信息查询，还能与车主进行自然流畅的语音对话。不过，这项AI功能需要用户激活高级车载娱乐服务才能使用。

ByteDance's Seed team has unveiled Seedance 2.5, a video generation model that creates up to 30-second single-take clips with enhanced realism and control. It supports multiple images, videos, and audio as references, plus precise editing via timestamps. Available on Jimeng AI and Doubao, it's set to transform filmmaking, ads, and more.

DeepSeek-V4-Flash API正式开启公测，标志着其Agent能力全面升级。新版本在推理速度、多模态处理及工具调用上均有显著提升，为开发者提供了更高效、更灵活的AI解决方案。本文将深入解析其核心特性、应用场景及对行业的影响。

China's AI sector is hitting new milestones, with domestic large models surpassing 10 billion downloads and open-source models reaching trillion-parameter scale. At a recent press conference, the government outlined plans to boost innovation, expand applications, and accelerate legislation for an Artificial Intelligence Law. The focus is on balancing rapid growth with safety and addressing impacts on jobs and the economy.

DeepSeek V4 正式版迟迟未发布，但多个线索指向8月3日这个关键节点。硅基流动平台宣布将于该日大幅上调 V4 Pro 的缓存命中价格，开发者社区也发现部分 API 已能回答此前无法回答的问题，暗示渐进式测试正在进行。与此同时，OpenAI 大幅降价，GLM 涨价，竞争格局愈发激烈。DeepSeek V4 能否在性能上扳回一城，答案或许就在8月3日揭晓。

WeChat Official Account has rolled out an AI-powered formatting feature that handles paragraphing, subheadings, and image alignment in one click. This tool aims to save creators time and effort, letting them focus on content rather than layout. With two styles and customizable colors, it's a handy addition for frequent publishers.

微软云业务年收入首次突破千亿美元，但增速放缓至43%，被谷歌云的82%增速反超。谷歌凭借自研芯片和模型，利润率大幅提升，而微软仍依赖英伟达芯片，利润结构承压。此外，微软678亿美元的合同积压中，很大部分依赖单一客户，引发市场对基础设施风险的担忧。在激烈的算力军备竞赛中，如何平衡增长与财务健康成为关键。

In his final earnings call as Apple CEO, Tim Cook hinted at potential Chinese memory chip suppliers amid a supply crunch, and confirmed Apple Intelligence has been approved for launch in China. With revenue beating expectations but stock falling on weak guidance, Cook leaves his successor a mixed legacy of challenges and opportunities.

A new report from Apollo Global Management reveals a surprising twist in the AI job debate: while AI hasn't caused widespread unemployment, it's quietly suppressing wage growth. Since 2023, workers in AI-exposed roles have seen their paychecks shrink, with low-income earners hit hardest. Even programmers aren't immune. But some professions, like financial advisors, are thriving. This isn't about losing your job—it's about losing your leverage.

DeepMind has unveiled Gemini Robotics ER 2, an AI model that lets robots plan and act in real time, and for the first time, collaborate with each other. This could speed up the use of robots in warehouses and factories, making them more efficient and adaptable.

Anthropic is making a massive bet on AI infrastructure with a new $15 billion data center in Texas, and Google is stepping in as a key backer. The tech giant will provide debt guarantees and supply custom chips, deepening their partnership. This move highlights the growing financial interdependence between AI startups and cloud giants as computing demands soar.

A top AI hedge fund, once boasting a 439% annual return, has been forced to sell most of its public holdings to Citadel after a sharp downturn in AI infrastructure stocks. Despite the setback, founder Leopold Aschenbrenner remains bullish, keeping a $5 billion stake in Anthropic. Is this a strategic pause or a sign of deeper trouble?