Welcome to the "AI Daily" section! This is your guide to exploring the world of artificial intelligence every day. Each 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.
Hot AI products click to learn more:https://top.aibase.com/
1. Bilibili tests new AI video creation tool "Peanut AI", complete a video in 3 minutes
Bilibili is testing a new AI video creation tool called "Peanut AI," which can help users quickly generate video content with high-quality results. At the same time, Bilibili is also advancing the application of AI technology, such as its self-developed large language model and multilingual translation features.
【AiBase summary:】
🎥 Peanut AI can quickly generate video clips from text or audio.
🔄 Users can choose to match materials intelligently or use templates for video creation.
🌐 Bilibili has launched a self-developed large language model that supports real-time multilingual translation.
2. Kunlun Wanzhi open sources the "Skywork UniPic 2.0" model
Kunlun Wanzhi announced the open sourcing of the "Skywork UniPic 2.0" model during the SkyWork AI technology release week. This unified multimodal model achieves efficient, high-quality, and unified multimodal generation capabilities through lightweight design of generation and editing modules, as well as joint training of multimodal understanding models.
【AiBase summary:】
🧠 Skywork UniPic 2.0 is a unified multimodal model that supports text and image input, expanding image generation and editing capabilities.
🚀 Based on the SD3.5-Medium architecture, it achieves integrated understanding, generation, and editing capabilities through joint training.
💡 Reinforcement learning is introduced, using a progressive dual-task strategy to improve the model's understanding of complex instructions and image generation consistency.
Details link: https://unipic-v2.github.io/
3. Alibaba Tongyi Qianwen officially launches Qwen Code!
Alibaba Cloud's Tongyi Qianwen launched the AI coding agent Qwen Code, providing developers with powerful and easy-to-use tools, while reducing costs and promoting the popularity of AI in the software development field.
【AiBase summary:】
🤖 Qwen Code is an AI coding agent based on the command-line interface, supporting tasks such as code generation, completion, debugging, and explanation.
🌐 Alibaba Cloud Tongyi Qianwen provides daily 2000 free API calls for mainland users, and 1000 for overseas users.
🚀 Qwen Code is based on the Qwen3-Coder series model, with strong technical support and performance comparable to industry-leading models.
4. Musk accuses Apple of favoring OpenAI, Apple responds strongly!
Musk accused Apple of favoring OpenAI in the App Store, while Apple responded that the platform design is fair and impartial. The conflict between the two sides is escalating, and they may face legal and market battles in the future.
【AiBase summary:】
🍎 Musk accused Apple of favoring OpenAI in the App Store, affecting the development of other AI companies.
⚖️ Apple emphasized that the App Store design is fair and impartial, ensuring developers get exposure opportunities.
🔥 Musk threatened to take legal action against Apple, and the conflict between the two sides escalated.
5. Tencent Hunyuan releases the 52B parameter multimodal understanding model Large-Vision, supporting arbitrary resolution full-scenario input
Tencent Hunyuan team released a new multimodal understanding model - Hunyuan Large-Vision. This model uses the MoE architecture, with a parameter scale of 52B, supporting arbitrary resolution images, videos, and 3D space input, while improving the ability to understand multilingual scenarios, providing a strong technical foundation for AI visual understanding services.
【AiBase summary:】
🖼️ Supports arbitrary resolution image processing, without complex preprocessing operations.
🧠 Uses MoE architecture to achieve a balance between performance and efficiency.
🌐 Improves multilingual scenario understanding capability, supporting cross-language multimodal applications.
6. Officially denies DeepSeek-R2 model's August release plan
DeepSeek officially denied rumors that its next-generation large model, DeepSeek-R2, would be released between August 15th and 30th. Although there were previous reports that R2 model would be released in March or May, all were denied by the official. Currently, DeepSeek has not yet announced the specific release date and technical details of the R2 model.
【AiBase summary:】
📌 DeepSeek officially denied the rumor that R2 model would be released in August.
📌 The release date and details of the R2 model have not been officially announced yet.
📌 DeepSeek team is still working on performance improvements and has not yet prepared for official deployment.
7. OpenAI launches affordable ChatGPT Go package: only 399 rupees, global users envy
OpenAI launched the ChatGPT Go package at an extremely low price (399 Indian rupees), providing users with more AI service functions, including longer conversation support, file upload, and image generation, aiming to reduce the usage threshold and attract more users.
【AiBase summary:】
🤖 ChatGPT Go package offers more conversation counts and file upload features.
🎨 Users can enjoy stronger image generation capabilities.
📊 Go package includes advanced data analysis features.
8. AI upstart Perplexity splurges $34.5 billion, officially offers to acquire Google Chrome
Perplexity proposed a cash offer of $34.5 billion to acquire Google Chrome browser, aiming to address Google's antitrust issues. It promised to keep Chromium open source and maintain Chrome's default settings, while demonstrating a deep understanding of the market and competitive landscape.
【AiBase summary:】
💰 Perplexity offered $34.5 billion in cash to acquire Chrome, showing strong ambition.
🔒 Promised to keep Chromium open source and maintain Chrome's default settings, reducing regulatory concerns.
🌐 Chrome holds a 68% market share, which could spark an intense bidding war.
9. Anthropic's Claude Sonnet 4 model now supports up to 1 million tokens
Anthropic launched the Claude Sonnet 4 model, extending its context token support to 1 million, significantly enhancing development flexibility. At the same time, it introduced a new pricing plan and put pressure on Google Gemini.
【AiBase summary:】
🤖 Anthropic's Claude Sonnet 4 model supports up to 1 million context tokens, enhancing development flexibility.
💰 New pricing plans are differentiated based on different token ranges, allowing developers to reduce costs through batch processing.
🔄 OpenAI is watching the long-context demand, but faces competition pressure from Anthropic and Google Gemini.
10. ChatGPT major update! GPT-5 mode switching freely, 4o back, experience more warm
OpenAI made a significant upgrade to ChatGPT, restoring GPT-4o as the default model and introducing multiple modes for GPT-5. At the same time, it optimized the AI's interactive personality, making it warmer and more user-friendly, thus enhancing the overall user experience.
【AiBase summary:】
✨ GPT-5 supports three mode switches, meeting different scenario needs.
🔄 Model selector optimization, paying users can access various models flexibly.
💬 AI interaction personality upgrade, improving user experience and naturalness.