Welcome to the 【AI Daily】 column! Here is your guide to exploring the world of artificial intelligence every day. We present you with the latest trends and highlights in the AI field, focusing on developers to help you gain insights into technological trends, and learn about innovative AI product applications.
Fresh AI products click to learn more: https://top.aibase.com/
1. Bilibili team launches AniSora open-source anime video generation model for creating multiple styles of animations with one click!
The Bilibili team has launched the AniSora open-source anime video generation model, filling the technical gap in the field of anime video generation. It supports the creation of videos in various anime styles.
[AiBase Summary:]
😊 AniSora supports the generation of multiple anime style videos with one click, covering series episodes, Chinese original animations, and more types.
🌟 Introduces a temporal mask module, supporting image-to-video generation, frame interpolation, and local image-guided features, improving the quality of generation.
🏆 After rigorous testing, person-motion consistency reached the current highest standard (SOTA), showcasing excellent performance.
2. OpenAI releases new programming assistant Codex
As a developer, I am very excited about OpenAI's release of Codex. Codex not only significantly shortens development time but also integrates seamlessly with GitHub, greatly enhancing work efficiency. It generates code that conforms to human preferences through reinforcement learning, demonstrating strong self-delegation capabilities.
[AiBase Summary:]
🚀 Codex intelligent assistant released by OpenAI can complete complex development tasks within 30 minutes.
🔗 Codex integrates seamlessly with GitHub, supporting multitasking processing, greatly improving developer efficiency.
🤖 Codex is trained through reinforcement learning to ensure the generated code conforms to human developer preferences.
3. Google Search introduces AI Mode experiment, exploring a new intelligent question-and-answer experience
Google has launched an experimental feature called 'AI Mode', providing an intelligent question-and-answer experience for text, voice, and image inquiries, and encourages user feedback to continuously optimize services.
[AiBase Summary:]
🌟 Supports text, voice, and image inquiries, providing a smarter question-and-answer experience.
🔍 Allows for in-depth exploration through follow-up questions, gaining more relevant information and web links.
🔒 Focuses on user privacy, takes measures to protect data security, and encourages user feedback.
Details link: https://support.google.com/websearch/answer/16011537?visit_id=638832352895396136-3267382421&p=aimodeavailability&rd=1#aimodeavailability
4. ChatGPT will integrate MCP protocol to help enterprises connect with diverse AI services
ChatGPT will soon support the MCP protocol, which allows it to seamlessly connect with third-party AI services, providing a more personalized user experience. Enterprises can optimize workflows through this protocol, enhancing efficiency and decision-making quality.
[AiBase Summary:]
🌟 The MCP protocol aims to unify the interaction methods between large language models and external systems, similar to the "USB-C interface" for AI applications.
⚙️ Users can customize adding tools, filling in names, URLs, and descriptions to combine ChatGPT with personal applications.
💼 The MCP provides enterprises with data-on-demand sharing capabilities, optimizing workflows, and promoting intelligent decision-making.
5. Alibaba Tongyi Lab launches ZeroSearch: Enabling large models to "search" without APIs
ZeroSearch is a new framework that enables large language models to simulate search engines, improving retrieval and reasoning capabilities while reducing reliance on real search engines and lowering training costs. This is achieved through reinforcement learning and a small amount of labeled data.
[AiBase Summary:]
✨ ZeroSearch uses reinforcement learning and a small amount of labeled data, enabling large models to generate high-quality documents without relying on real search engines, enhancing reasoning capabilities.
📚 The framework adopts a curriculum-based learning method, gradually training from high-quality to low-quality documents, improving the model's ability to handle complex retrieval tasks.
🌟 In question-answering datasets, ZeroSearch outperforms traditional methods, showing significant advantages in single-hop and multi-hop question answering tasks.
6. Stability AI and Arm release mobile-grade audio generation AI: Creates 11 seconds of stereo sound in 7 seconds
Stability AI and Arm jointly released a stable audio open small model that can generate 11 seconds of high-quality stereo audio in 7 seconds, optimized to run smoothly on mobile devices. Based on adversarial relative contrast technology, it significantly reduces parameter quantities, making it suitable for consumer-grade hardware.
[AiBase Summary:]
Breakthrough technology allows audio generation in just 7 seconds, achieving near-real-time audio synthesis capabilities.
The model architecture is optimized into three parts, adapted for mobile use, supporting various audio generation tasks.
Training data is strictly selected to ensure legality and compliance, but is currently more suitable for English prompt input.
7. Qwen releases new preference modeling model series WorldPM
The Qwen team has released the WorldPM series models, including WorldPM-72B and its derivatives, achieving breakthroughs in preference modeling through large-scale training, providing developers with efficient optimization paths.
[AiBase Summary:]
🌍 WorldPM is trained using 15 million pieces of preference data, verifying that preference modeling follows scaling laws, improving model performance in supervised learning.
🌐 The model series is open-sourced, reducing technical barriers, and helping global developers improve model optimization efficiency.
🌟 Enhances style neutrality, overcoming subjective biases, and shows significant advantages in coding and mathematical tasks.
Details link: https://huggingface.co/Qwen/WorldPM-72B
8. OpenAI unveils GPT-5: Integrating multiple products into one
Jerry Tworek shared the latest updates on GPT-5 on Reddit, stating that it will integrate Codex, Operator, Deep Research, and Memory to simplify user workflows. Codex’s programming efficiency has tripled, and OpenAI plans to use this tool to help novice developers get started faster.
[AiBase Summary:]
🌟 GPT-5 integrates Codex, Operator, Deep Research, and Memory, reducing the hassle of switching between tools.
💻 Codex triples programming efficiency, especially suitable for developers solving trivial problems.
👨💻 OpenAI plans to use Codex to help novice developers quickly learn programming, enhancing the overall capability of human developers.
9. ListenHub: AI-generated tool officially launches, revolutionizing podcast experiences
ListenHub is an AI-based podcast generation tool that supports both Chinese and English, offering personalized podcast experiences. It is popular due to its efficient generation speed and user-friendly interface, suitable for ordinary users and content creators. It offers free and premium membership services and focuses on mobile experiences.
[AiBase Summary:]
🌟 Uses AI technology to quickly generate content related to user interests, covering topics such as technology, history, and society.
⚡️ Fast generation speed, completing podcast production in 1-5 minutes, suitable for busy people and content creators.
📱 Supports multiple platforms and mobile use, offering free and premium membership options to meet diverse needs.
Details link: https://listenhub.ai/zh
10. QQ Browser upgrades to AI Browser: Launches QBot with five new AI capabilities
QQ Browser has been upgraded to an AI browser and launched QBot, bringing a smarter browsing experience, including search, reading, translation, writing, and office assistance functions.
[AiBase Summary:]
🚀 QBot supports multimodal questioning and can accurately answer various questions, providing 24/7 smart companionship.
📚 The AI reading tool can quickly summarize web page content, generate mind maps, and improve information processing efficiency.
💼 In office scenarios, QBot provides document editing, translation, writing, and other multifunctional tools, assisting efficient office work.
11. MathModelAgent: An AI assistant for mathematical modeling
MathModelAgent is an intelligent tool specifically designed for mathematical modeling, capable of automatically completing the entire process from problem analysis, model building, code writing, to paper writing. It demonstrates the profound potential of AI in academic and technical fields.
[AiBase Summary:]
Problem Analysis and Modeling: The modeling hand can quickly parse mathematical problems and generate logically clear mathematical models.
Code Generation and Debugging: The code hand includes a reflection module, generating high-quality code and debugging it in real time through a local interpreter.
Paper Auto-Writing: The paper hand automatically generates academically formatted papers based on the results of modeling and computation.
12. GenSpark releases the world's first Agentic AI download agent, revolutionizing file management experience
I am very optimistic about GenSpark's release of this Agentic Download Agent tool. It truly achieves automation and intelligence in file management and information processing, greatly simplifying my workflow. Whether for academic research or daily office work, this tool allows me to focus on more important things.
[AiBase Summary:]
🚀 Supports one-click completion of file search, download, and organization through natural language instructions, greatly increasing efficiency.
📚 Provides AI Drive functionality, supporting document summarization, key information extraction, and analysis report generation.
🌐 Features powerful automation and intelligence, supporting batch processing, intelligent organization, and transparent operations.
13. Google NotebookLM is about to launch Sparks video overview
Google's NotebookLM plans to launch the 'Sparks' feature, converting documents, notes, etc., into 1-3 minute videos, 10% of which are generated by AI. Combining Gemini2.5 and Deep Research functions, it provides an end-to-end solution from research to presentation.
[AiBase Summary:]
✨ Sparks video overview combines Gemini2.5 and Deep Research, converting documents into 1-3 minute videos, aiding efficient content creation.
📚 Suitable for education, research, content creation, and more scenarios, significantly improving work efficiency.
🌐 Globally deployed, supporting multiple languages, with potential for further international expansion.