Welcome to the "AI Daily" section! This is your guide to exploring the world of artificial intelligence every day. Every day, we present you with the latest content in the AI field, focusing on developers, helping you understand technology trends and learn about innovative AI product applications.
New AI products click to learn more:https://app.aibase.com/en
1、MiniMax's Hailuo AI first and last frame feature officially launched on web and app
The first and last frame feature of Hailuo AI has been officially launched on both the web version and the APP, and it also opens up only the last frame play mode. This technology comprehensively improves the upper limit of the industry's first and last frame capabilities through stronger instruction understanding, smoother dynamic effects, and bolder imagination.
【AiBase Highlights:】
🧠 Strongest complex instruction following capability, accurately understanding and executing every detail
🎬 Extreme complex physical dynamic generation, such as smooth combo moves for high-energy actions like fighting and gymnastics
🎨 Unexpected imagination function, achieving out-of-the-box performance when there is a large gap between the first and last frames or lack of instructions
2、Yuan Shi Technology launches Wen Xiao Bai 5, challenging GPT-5, a new benchmark for domestic AI is coming
The latest flagship product of Yuan Shi Technology, Wen Xiao Bai 5, is close to GPT-5 in multiple performance tests, marking an important breakthrough in domestic large model technology. The system has a dynamic thinking mode, applicable to multiple fields, and performs well in STEM capabilities, cutting-edge knowledge, and code programming.
【AiBase Highlights:】
✨ Wen Xiao Bai 5 is close to GPT-5 in multiple performance tests, becoming a new benchmark for domestic AI.
🧠 It has a dynamic thinking mode, intelligently judging when to respond quickly or think deeply.
📊 It performs excellently in STEM capabilities, cutting-edge knowledge, and code programming, with a comprehensive score exceeding similar products.
3、OpenAI releases a new voice model GPT-Realtime, designed for voice AI Agents
OpenAI released a new voice model called GPT-Realtime, specifically designed for voice AI agents. It can generate natural and fluent voice and supports image input and multilingual switching. It has significant improvements in reasoning ability and instruction following accuracy, while providing powerful security protection features suitable for multiple industries.
【AiBase Highlights:】
🎙️ GPT-Realtime is a multimodal voice model released by OpenAI, specifically designed for voice AI agents.
🧠 The model has reasoning and instruction following capabilities, improving the level of intelligent voice interaction.
🔒 The Realtime API is equipped with security protection measures to ensure user privacy and data security.
4、Say goodbye to complicated! Google Gemini AI makes table processing effortless
Google introduced the Gemini AI assistant, making data processing in Google Sheets more intelligent and efficient, enhancing the user experience.
【AiBase Highlights:】
📊 Google Gemini AI assistant brings intelligent data processing functions to Google Sheets.
💡 The new "Convert to Table" feature automatically analyzes and organizes data, improving work efficiency.
🔄 Users can customize formula expressions to adapt to data changes without manually adjusting formulas.
5、AI Voice Acting Revolution! Tencent's Black Tech Makes Machines Become Top Storytellers, Generating Hollywood-Level Sound Effects with One Sentence
The article introduces the AudioStory technology developed by Tencent ARC Lab, which can generate high-quality audio content based on text descriptions and has strong narrative capabilities. It realizes complex audio generation tasks through a divide-and-conquer strategy and a decoupled connection mechanism.
【AiBase Highlights:】
✨ AudioStory technology can generate movie-level audio content based on text descriptions.
🧠 It uses a divide-and-conquer strategy to break down complex stories into ordered audio events.
🔄 A decoupled connection mechanism ensures precise matching of audio quality and semantics.
Details link: https://arxiv.org/pdf/2508.20088
6、Baidu plans to cultivate 10 million AI talents in the next five years
The article introduces Baidu's plan to cultivate 10 million AI talents in the next five years, while showing its continuous investment and innovation achievements in the field of artificial intelligence. Additionally, the article mentions that Baidu's new business revenue from AI has performed outstandingly, showing its competitiveness in the market.
【AiBase Highlights:】
🌟 Baidu plans to cultivate another 10 million AI talents in the next five years, promoting the development of the industry.
📈 Baidu's Q2 2025 financial report shows that the revenue from its new AI business exceeded 10 billion yuan, with a year-on-year growth of 34%.
🎓 Talent cultivation will be carried out through various methods such as university cooperation, enterprise training, and online education.
7、Anti-cheating AI tutor emerges! MathGPT.ai successfully pilot tested at 30 US universities, to be widely promoted this fall
MathGPT.ai redefines the role of AI in mathematical education through Socratic teaching methods and teacher-led control mechanisms. The platform not only provides anti-cheating tutor services but also supports university-level math courses and integrates with mainstream learning management systems to ensure seamless access.
【AiBase Highlights:】
🧠 MathGPT.ai uses Socratic questioning technology to encourage students to think critically rather than directly obtaining answers.
🔒 Teachers can control how students use AI tools, including specifying whether AI provides tutoring support.
🌐 The platform is integrated with Canvas, Blackboard, and Brightspace, and is compatible with screen readers, enhancing accessibility experiences.
8、Apple Xcode重磅集成Claude Sonnet4:iOS开发迎来AI革命时代
Apple officially integrated the Claude Sonnet4 AI model into Xcode 26 Beta 7, bringing an intelligent programming experience to iOS developers. The model can generate high-quality code, locate errors, and automatically fix them. The newly added inline playgrounds feature allows developers to run and test code directly in the code line, improving development efficiency.
【AiBase Highlights:】
🍎 Integrated with the Claude Sonnet4 AI model, enhancing code generation and error correction capabilities.
🧪 New inline playgrounds feature supports real-time execution of code examples.
🔒 Implemented based on Apple's official extension interface, ensuring stability and security of the functionality.
9、Microsoft launches its first self-developed AI models MAI-Voice-1 and MAI-1-preview, competing with OpenAI
Microsoft has launched its first self-developed AI models MAI-Voice-1 and MAI-1-preview, marking an important advancement in the field of artificial intelligence and strengthening its competitiveness against OpenAI. MAI-Voice-1 can quickly generate audio and has been applied to functions such as Copilot Daily; while MAI-1-preview focuses on daily query assistance and will be used for text processing in the Copilot AI assistant in the future.
【AiBase Highlights:】
🗣️ MAI-Voice-1 can quickly generate audio and has been applied to multiple functions such as Copilot Daily.
🚀 MAI-1-preview will be used for text processing in the Copilot AI assistant, marking Microsoft's new progress in the consumer-grade AI field.
🌟 Microsoft has launched two self-developed AI models, MAI-Voice-1 and MAI-1-preview, enhancing its competitiveness against OpenAI.
Details link: https://microsoft.ai/news/two-new-in-house-models/
10、xAI隆重推出Grok Code Fast1:快速、经济的高效代理编码模型
xAI launched Grok Code Fast1, a large language model designed for software development that is fast and cost-effective. The model performs well in reasoning and code generation capabilities and has been freely available on multiple mainstream intelligent programming platforms.
【AiBase Highlights:】
🚀 Grok Code Fast1 adopts a new lightweight model architecture, improving service speed and cache hit rate.
🌐 Supports multiple platforms such as GitHub Copilot and Cursor, offering free trials to attract developers to experience.
💰 Competitive pricing strategy, at $0.20 per million input tokens and $1.50 per output token, suitable for budget-conscious developers.
Details link: https://x.ai/news/grok-code-fast-1
11、SuperCLUE多模态视觉8月评测榜：Gemini-2.5-Pro位居第一
In the SuperCLUE-VLM benchmark list released on August 28, Gemini-2.5-Pro ranked first with 74.99 points, followed by GPT-5(high) from OpenAI with 68.59 points. The list was built around three dimensions: basic cognition, visual reasoning, and visual application, aiming to provide an objective and fair evaluation standard for multimodal vision language models.
【AiBase Highlights:】
🧠 Gemini-2.5-Pro ranked first in the SuperCLUE-VLM list with 74.99 points, demonstrating strong multimodal capabilities.
📊 The evaluation covers 15 multimodal models, including Claude-Opus-4.1, GPT-5(high), and other domestic and international mainstream models.
🏆 Baidu ERNIE-4.5-Turbo-VL ranks alongside other domestic models, showing strong market competitiveness.
12、AI Content Labeling Regulations on September 1st! Non-compliance Directly Bear Legal Risks, Professionals Should Quickly View the Risk Avoidance Guide
The article details the implementation background and core requirements of the national standard GB45438-2025, "Methods for Identifying AI-Generated Synthetic Content." The standard clearly defines the labeling methods for AI-generated content, the identification of responsible entities, and the consequences of violations, emphasizing the importance of AI content governance.
【AiBase Highlights:】
📌 Explicit labeling requires AI-generated content in different forms such as text, images, and videos to be clearly marked with AI attributes.
🔍 Implicit labeling embeds AIGC identifiers in file metadata to ensure traceability of content sources.
⚖️ Violations have serious consequences, including traffic restrictions, rectification, removal, and legal risks. Enterprises need to immediately prepare for compliance.