Discover the AI Tools

Lingguang App's Flash App platform has surpassed 4 million creators, most without programming backgrounds. From students to therapists, these everyday users are building AI applications for fun, emotional support, and practical decisions. Simulators lead the trend, with nearly 10,000 apps, including an 'Idol Simulator' with 13.35 million interactions. This shift marks AI's evolution from a professional tool to a creative playground for all.

Google has temporarily pulled the plug on its new AI image generation feature in Google Earth, less than two days after it went live. The tool, powered by the Nano Banana2 model, lets users create fictional scenes on real satellite imagery—but tests showed it could produce convincing fake disasters, raising fears of misinformation. Google says it's strengthening security measures, but the incident highlights the growing challenge of keeping AI creative tools from becoming tools of deception.

AI安全评估机构METR面临严峻的人才短缺问题，尽管提供高达50万美元的年薪，仍难以招到足够的研究人员。该机构由前OpenAI研究员创立，负责评估各大AI实验室模型的风险，但团队仅35人，工作量远超负荷。随着AI能力每七个月翻倍，安全评估人才的增长却远远跟不上，这一缺口正成为AI时代最危险的结构性风险之一。

8月3日，MiniMax开源了其首款多模态生成模型H3，摩尔线程当天便完成了适配。借助MTT S5000智能计算卡和MUSA软件栈，H3在3小时内实现快速部署。该模型支持文本、图像、音频和视频的全面输入，能直接生成2K分辨率、最长15秒的影音内容，视频生成成本低至每秒0.8元，商业性价比高。此次合作展示了国产算力生态的响应速度，也为多模态AI的普及铺平了道路。

Alibaba has officially launched Qwen3.8-Max, a massive AI model with 2.4 trillion parameters, showing major gains in coding and office tasks. It ranks second only to Anthropic's Claude on the Arena leaderboard. The API is already live, and the model, along with Qwen3.8-27B, is expected to be open-sourced next week, signaling a strong push for open-source AI.

On August 3, MiniMax unveiled its next-generation video model, H3, as an open-source release. This multimodal system can handle text, images, video, and audio together, generating up to 15-second clips at 24 FPS with stereo sound. With a special 2K mode, it delivers stunning resolution. Supporting 11 languages and flexible input options, H3 aims to democratize AI video creation, making professional-grade tools accessible to everyone.

West Lake Xinchen, a prominent Chinese AI startup, has secured a substantial B+ funding round worth hundreds of millions of yuan. The investment, led by Guangfa Xinde, includes participation from major financial and industrial players like Ant Group and 58 Industrial Fund. The fresh capital will fuel next-gen diffusion language models, enhance multimodal real-time interactive products, and accelerate global expansion, positioning the company to compete on the world stage.

据知情人士透露，蚂蚁集团旗下的具身智能公司蚂蚁灵波科技可能启动独立融资，这意味着这家由蚂蚁集团全资孵化的企业将正式走向外部资本市场，接受市场考验。随着集团在AI领域的投入不断加大，内部资源分配压力上升，引入外部资本成为支持后续研发和业务扩张的关键一步。目前，具身智能赛道融资环境活跃，蚂蚁灵波的估值有望达到独角兽级别。此外，蚂蚁集团旗下蚂蚁数科也在筹备Pre-IPO融资，加上此前已启动独立融资的蚂蚁国际和OceanBase，已有四家子公司踏上资本化进程。

IBM与Ponemon研究所的最新报告揭示了一个令人担忧的趋势：AI驱动的攻击已占所有恶意数据泄露的四分之一，同比增长56%，每次事件平均损失达600万美元，比整体平均水平高出20%。更令人警惕的是，62%的AI攻击瞄准关键基础设施，金融和能源领域首当其冲。报告还指出，攻防成本严重失衡——攻击更便宜，防御更昂贵，而多数企业尚未充分利用AI自动化来应对这一挑战。

腾讯元宝AI助手迎来重大更新，图像理解能力全面升级，推出“先别慌，拍个照”新场景。现在，拍张照片就能读懂体检报告、解析租房合同、排查家电故障，甚至帮你回复聊天消息。这项功能覆盖医疗、生活、社交等多个日常领域，让AI真正成为解决生活小麻烦的贴心助手。

Elon Musk showcased Grok's latest video interaction feature by feeding it an AI-generated video of Kobe Bryant. The model not only described the visuals and dialogue but also correctly identified the footage as a deepfake, highlighting its advanced reasoning abilities. This leap from audio summaries to full video understanding marks a new era in AI interaction.

韩国国家AI计算中心在罗州破土动工，总投资2.5万亿韩元，预计2028年建成。三星SDS领投，政府与民企深度合作，旨在打造全球高性价比算力枢纽，并推动国产NPU研发。这一大手笔，能否让韩国在全球AI竞赛中抢占先机？

DeepSeek发布了V4-Flash稳定版，并同步开放API公测。国家超算互联网平台已上线该模型的API调用和文件下载服务。新版模型在智能体能力和指令遵循上大幅提升，多项基准测试成绩已与当前最强闭源模型持平。开发者无需复杂配置，登录平台即可一键调用，还能下载模型文件进行微调和部署。目前平台已汇聚超1700个开源大模型，为AI应用开发提供了便捷的试验场。

AI startup Moonlight has reportedly completed a massive $3.5 billion funding round and is now preparing for a Hong Kong IPO. The company, known for its Kimi model, is valued at $35 billion post-money, with plans to raise around $3 billion from the listing. This move highlights the accelerating commercialization of China's AI sector.

OpenAI's new model, Astra, has been turning heads with its math prowess, but not everyone is convinced. Renowned scholar Gary Marcus warns that the hype is a classic case of the 'fallacy of composition'—just because it's great at math doesn't mean it's ready for the real world. With limited details released, experts urge caution before we crown it the next AGI.

A Munich court has ruled that AI music generator Suno infringed copyright during both training and output, rejecting its fair use defense. The case, brought by German music rights organization GEMA, centered on songs like "Atemlos durch die Nacht" and "Rasputin." The court found Suno's models could reproduce original elements, a phenomenon known as "memorization." This decision could have far-reaching implications for AI training data practices.

SenseTime has just dropped the preview of its SenseNova U1.5-Lite-Preview, a compact 8B model that can natively generate 4K images. It's a significant upgrade from its predecessor, offering better texture details, more accurate text rendering, and more stable editing. The best part? It's open-source and available on GitHub, Hugging Face, and ModelScope, making high-end AI image tools accessible to everyone.

腾讯CodeBuddy宣布全面支持DeepSeek-V4-Flash正式版，旗下WorkBuddy、CodeBuddy IDE、插件及CLI产品均已适配。新模型在代码生成、任务执行等场景表现优异，显著超越前代版本，为开发者带来更高效的智能编程体验。

Moonshot AI has officially denied reports of filing for a Hong Kong IPO this month, but the company's recent actions tell a different story. With a massive $3.5 billion funding round and a corporate restructuring, the AI unicorn is clearly preparing for something big. What's next for this industry leader?

北京一家名为AGI Bar的酒吧推出创新服务：顾客消费后连接Wi-Fi即可免费获得DeepSeek V4 Flash的API Token，在店内无限量使用AI算力。这一模式将AI服务从付费订阅转变为像水电一样的线下基础设施，引发行业关注。