Welcome to the "AI Daily" column! This is your guide to exploring the world of artificial intelligence every day. Every day, we bring you the latest content in the AI field, focusing on developers to help you understand technical trends and innovative AI product applications.
Fresh AI products click to learn more:https://top.aibase.com/
1、Coze Space Web Design Feature Launches
Coze Space (coze.cn) has launched a web design feature that uses AI technology to shorten web design time from days to just 5 minutes, greatly improving design efficiency and lowering the design barrier. Users only need to input their requirements, and the system can generate a web page that matches the description, supporting natural language input and secondary editing.
AiBase Summary:
🌟 Coze Space realizes fast web design through AI technology, improves efficiency and lowers the design barrier.
🎨 Users can generate personalized web pages through natural language input or by uploading reference images.
🌐 This feature applies to various scenarios such as event marketing pages, institutional homepages, and personal homepages.
2、Qwen-MT, a Machine Translation Model Based on Qwen 3, is Launched by Tongyi Qianwen
Qwen-MT is a machine translation model developed based on the Qwen3 model, supporting bidirectional translation among 92 languages, with advantages such as high controllability, low latency, and low cost. It performs excellently in both automatic and manual evaluations, demonstrating outstanding translation capabilities.
AiBase Summary:
🌍 Supports bidirectional translation among 92 languages, covering over 95% of the global population.
⚙️ Provides professional translation functions such as terminology intervention, domain prompts, and memory libraries.
⚡ Lightweight MoE architecture enables fast response and low-cost API calls.
Details: https://bailian.console.aliyun.com/?tab=model#/model-market/detail/qwen-mt-turbo
3、ChatGPT Agent Function Fully Launched, Plus, Pro, and Team Users Can Now Experience
The launch of the ChatGPT Agent function marks a major advancement in the field of AI for task automation, providing users with a more efficient and accurate intelligent assistant experience.
AiBase Summary:
🤖 The ChatGPT agent function is fully launched, enhancing task automation capabilities.
📊 Performs well in multiple benchmark tests, significantly improving efficiency and accuracy.
🔒 Security has been enhanced, but financial operations still require user control.
4、Alibaba's Wan 2.2 is About to Shockingly Launch: Open-Source Video Generation AI Challenges Sora
Alibaba Cloud announced that Wan 2.2 is about to be released as an upgraded version of Wan 2.1, achieving significant breakthroughs in performance, efficiency, and functionality, further optimizing video generation technology and enhancing the multimodal creation experience.
AiBase Summary:
🎥 New text-to-video (T2V) function added, supporting higher resolution and longer video generation.
🎨 Supports multilingual and style expansion, adding cyberpunk, realistic animation, and other art style templates.
⚙️ Optimized hardware requirements, T2V-1.3B model can run on devices with low VRAM.
5、Anthropic Launches Audit Agent to Help Test AI Model Alignment
Anthropic has launched a new audit agent to improve the efficiency of AI model alignment testing. This technology is used to test the Claude Opus4 model before deployment, aiming to address the issue of AI models possibly overly catering to users. The research team developed three audit agents and open-sourced the code to encourage more researchers to participate.
AiBase Summary:
🔍 Audit agents are used to detect alignment issues in AI models, improving testing efficiency.
⚙️ Three audit agents are provided, each responsible for investigation, evaluation, and red team testing.
🌐 Open-source code encourages more researchers to explore and improve.
6、OpenAI to Release GPT-5, Expected to Launch in August
OpenAI's next-generation language model, GPT-5, is expected to officially launch at the beginning of August. CEO Sam Altman revealed that GPT-5 is progressing smoothly and mentioned its powerful reasoning capabilities, which are surprising. In addition, OpenAI plans to release an open-weight language model by the end of July, further promoting the popularization of AI technology.
AiBase Summary:
🌟 GPT-5 is expected to be released in August, integrating various reasoning capabilities, and the user experience will be significantly improved.
🔍 Will release mini and nano versions, expanding the application range of OpenAI tools.
📈 OpenAI plans to release an open-weight language model with advanced reasoning capabilities by the end of July.
7、Google Launches Opal: Build AI Applications Without Code Using Natural Language
Google Lab launched Opal, a no-code AI application development tool, allowing users to create AI-driven mini applications through natural language descriptions without programming knowledge.
AiBase Summary:
🧪 Converts natural language into visual AI workflows, simplifying the development process.
🚀 Supported by the Gemini model for rapid AI application generation, improving efficiency.
🌐 Supports cloud sharing, promoting collaboration and innovation.
8、Nanyang Technological University and Shanghai AI Lab Launch PhysX-3D: Injecting 'Physical Soul' into AI-Generated 3D Models!
The article discusses the current problem of AI-generated 3D models lacking physical properties and introduces the PhysX-3D project launched by Nanyang Technological University and the Shanghai AI Lab. This project provides a new method for AI-generated 3D models with real physical properties by constructing the PhysXNet dataset and developing the PhysXGen generation framework.
AiBase Summary:
📌 The PhysX-3D project aims to solve the problem of AI-generated 3D models lacking physical properties.
💡 Proposed the 'Five Questions of the Soul' for 3D models, covering core dimensions such as size, material, and functional affordance.
🚀 The PhysXGen generation framework combines geometric and physical properties to achieve more realistic 3D modeling.
Details: https://arxiv.org/pdf/2507.12465
9、Google Lab's Groundbreaking Product Opal: No Code! Create AI Applications with Natural Language, Unlock Future Productivity
Google Lab's Opal is a revolutionary experimental AI tool that allows users to quickly create AI-driven mini applications through natural language processing and visual editing without any coding. Its core features include natural language-driven, visual workflow, integration with the Google AI ecosystem, and sharing and collaboration, providing a convenient AI development experience for developers and general users alike.
AiBase Summary:
✨ Opal allows users to describe their needs through natural language, automatically generating AI application logic.
🎨 Provides a visual workflow editor, allowing users to intuitively adjust application steps.
🌐 Integrates Google AI models (such as the Gemini series), enabling multimodal processing capabilities.
Details: https://developers.googleblog.com/en/introducing-opal/
10、Kuaishou Opens KAT-V1 Large Model: Significant Improvement in Autonomous Thinking Ability, 40B Version Performance Close to 40B
Kuaishou officially released and open-sourced the KAT-V1 autonomous thinking large model, which excels in the integration of thinking and non-thinking abilities, capable of adjusting modes based on the complexity of the question. The 40B version's performance is close to DeepSeek-R1, while the 200B version surpasses several flagship models in multiple benchmark tests.
AiBase Summary:
🧠 KAT-V1 has integrated autonomous thinking and non-thinking abilities, capable of adjusting modes based on task complexity.
🚀 The 40B version's performance is close to DeepSeek-R1, while the 200B version outperforms Qwen, DeepSeek, and Llama series in benchmark tests.
🛠️ Uses the reinforcement learning algorithm Step-SRPO to enhance reasoning ability and thinking density, optimizing excessive thinking problems.
Details: https://huggingface.co/Kwaipilot/KAT-V1-40B
11、iFlytek Spark X1 Deep Reasoning Large Model Upgrade Released
iFlytek launched the Spark X1 upgrade, a deep reasoning large model trained using domestically produced computing power, significantly enhancing comprehensive capabilities, with notable progress in hallucination management, multilingual support, and voice simultaneous interpretation, providing smarter, more reliable, and efficient AI solutions for multiple industries.
AiBase Summary:
✨ Spark X1 has made significant progress in hallucination management, improving the reliability of large models.
🌐 Multilingual support covers over 130 languages, enabling barrier-free cross-language communication.
🚀 Voice simultaneous interpretation technology has improved, with translation quality scores exceeding 90 and response times shortened to 2 seconds.
Details: https://xinghuo.xfyun.cn/desk