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, helping you understand technology trends and learn about innovative AI product applications.
Fresh AI products click to learn more:https://top.aibase.com/
1. Volcano Engine launched the latest Dabao image editing and simultaneous interpretation models, accelerating Agent development and implementation
Volcano Engine announced the new Dabao series models and AI cloud-native service upgrades at the Xiamen event, including image editing model 3.0, simultaneous interpretation model 2.0, and the large model 1.6 series. It also introduced the open-source core capabilities of Kousi and enterprise self-hosted model solutions, helping developers build Agents and implement AI applications.
【AiBase Summary:】
🔥 Dabao Image Editing Model 3.0 enhances natural language instruction processing capabilities, suitable for film creation and advertising marketing fields.
🎙️ Dabao Simultaneous Interpretation Model 2.0 achieves low-latency voice synchronization generation, supporting dialect and accent matching.
🚀 Dabao Large Model 1.6 Series optimizes code, reasoning, and math capabilities, reducing latency and cost.
2. Tongyi Qianwen opensource Qwen3-30B-A3B-Instruct-2507 non-thinking mode model
The new version of Qwen3-30B-A3B-Instruct-2507 performs excellently in non-thinking mode, with performance comparable to top closed-source models, achieving significant improvements in multiple areas, while supporting multilingual and long-text processing, providing a convenient open-source platform for developers and researchers.
【AiBase Summary:】
🧠 The new version of Qwen3-30B-A3B-Instruct-2507 activates only 3B parameters in non-thinking mode, with performance comparable to Gemini2.5-Flash(non-thinking) and GPT-4o, top closed-source models.
🌐 General capabilities have significantly improved, covering instruction following, logical reasoning, text understanding, mathematics, science, programming, and tool usage among other areas.
📚 Long-text understanding capability reaches 256K, with broader coverage of multilingual long-tail knowledge, suitable for global application scenarios.
3. OpenAI Launches New Learning Assistant ChatGPT Study
OpenAI has released a new ChatGPT Study learning mode, aimed at providing students and educators with a more personalized and interactive learning experience. This feature integrates interactive prompts, scaffolded responses, personalized education, and knowledge point checks, suitable for various subjects, and is available to all users.
【AiBase Summary:】
🚀 ChatGPT Study provides a personalized and interactive learning experience, suitable for students and educators.
📚 Integrates four main functions, including interactive prompts, scaffolded responses, personalized education, and knowledge point checks.
🌐 This feature is available to all free, Plus, Pro, and Team users, marking OpenAI's further advancement in educational technology.
4. China Launches HYPIR Image Restoration Large Model, Turning Old Photos into 8K High Definition in 1.7 Seconds
The launch of the HYPIR image restoration large model demonstrates China's innovation in the field of artificial intelligence and brings revolutionary changes to cultural heritage, film restoration, and scientific research and medical fields.
【AiBase Summary:】
🧠 HYPIR image restoration large model can restore old photos to 8K ultra-high-definition quality in 1.7 seconds.
📄 HYPIR excels in preserving text, capable of high-fidelity restoration of text content.
🚀 The release of HYPIR marks the beginning of a new era in image restoration technology.
5. Google Launches Video Overview Feature for NotebookLM, Enhancing Learning Experience
Google has launched the video overview feature for NotebookLM, converting complex information into slides with narration to provide users with a more intuitive learning experience. This feature can extract information from user-uploaded materials and plans to expand multilingual support in the future.
【AiBase Summary:】
📹 New feature launched: Google introduces the video overview feature for NotebookLM, adding an intuitive visual experience to learning.
📄 Data extraction capability: This feature can extract information from user-uploaded materials and automatically generate slides with narration.
🌍 Future expansion: The video overview feature will support multiple languages and will continue to enrich and expand its presentation forms.
6. Google Quietly Upgrades Imagen 4! Crushes GPT-4o, AI Image Generation King Returns?
Google has made a major upgrade to its text-to-image generation model Imagen4. After the upgrade, Imagen4Ultra climbed to third place on the authoritative Artificial Analysis image arena ranking, standing shoulder to shoulder with OpenAI's GPT-4o and Seedream3.0, becoming one of the world's top image generation models. It stands out in image quality, handling complex prompts, price advantage, and generation speed.
【AiBase Summary:】
🖼️ Imagen4Ultra excels in image details, realism, and style consistency, with significantly improved performance.
💰 The standard and Ultra versions of Imagen4 are much cheaper than GPT-4o, offering high cost-effectiveness.
⚡ On average, Imagen4Ultra generates an image in just 9.5 seconds, leading most competitors in efficiency.
7. Kunlun Wanhui Opensource Multimodal Unified Pre-training Model Skywork UniPic
Kunlun Wanhui's Skywork UniPic multimodal unified pre-training model combines image understanding, text-to-image generation, and image editing capabilities. Based on large-scale data end-to-end pre-training, it demonstrates excellent generality and transferability. The model performs outstandingly, especially running smoothly on consumer-grade GPUs, significantly lowering the technical application threshold.
【AiBase Summary:】
🧠 MAR encoder and SigLIP2 backbone network achieve deep collaboration across tasks.
🖼️ Supports complex operations such as image understanding, generation, and style transfer.
🚀 1.5B parameter scale approaches the performance of large unified models.
Details link: https://huggingface.co/Skywork/Skywork-UniPic-1.5B
8. Li Auto i8 Launch, First Equipped with Intelligent Driver Large Model, Starting Price 321,800 RMB!
Li Auto launched a new six-seat pure electric SUV - Li i8, offering three versions with starting prices of 321,800 RMB, 349,800 RMB, and 369,800 RMB. The car offers generous technical configurations, standard features include dual motor four-wheel drive system, self-developed 5C battery, and laser radar, and globally first equipped with VLA driver large model, enhancing vehicle safety and intelligence level.
【AiBase Summary:】
🚗 Li i8 is equipped with a dual motor four-wheel drive system, maximum power up to 400kW, torque of 660Nm.
🔋 Self-developed 5C battery provides two capacities, with range of 670 kilometers and 720 kilometers respectively.
🤖 Globally first equipped with VLA driver large model, supports natural language interaction, enhancing smart operation experience.
9. Google Launches AI Search Mode in the UK, Enhancing Complex Question Answering Experience
Google launched a new AI search mode in the UK, which helps users ask more complex questions and get accurate AI-driven replies. This mode is based on the Gemini2.5 model, uses query expansion technology to deeply explore web content, and supports multiple interaction methods, such as text, voice, or image questions.
【AiBase Summary:】
🧠 AI mode is based on Google's latest Gemini2.5 model, supporting complex multi-part question processing.
🔍 Query expansion technology decomposes user questions into multiple sub-topics, deeply exploring web content.
📱 Supports voice and image questions, improving interaction experience and convenience.
10. Kunlun Wanhui Opensource Multimodal Unified Pre-training Model Skywork UniPic, Integrating Image Generation and Understanding Capabilities
Kunlun Wanhui launched the open-source multimodal unified pre-training model Skywork UniPic, integrating image understanding, text-to-image generation, and image editing functions. The model is end-to-end pre-trained on large-scale high-quality data, showing good generality and transferability. Its 1.5B lightweight design approaches the performance of large models, and improves model performance through refined data construction and multi-stage training.
【AiBase Summary:】
📷 Skywork UniPic integrates image understanding, text-to-image generation, and image editing, achieving multimodal unification.
💡 Using a 1.5B lightweight parameter scale, the performance approaches that of large models, making it easy for developers to use.
🚀 Through multi-stage training and refined data construction, the model improves execution ability in instruction following and complex tasks.
Details link: https://huggingface.co/Skywork/Skywork-UniPic-1.5B
11. OWL team opensource a multi-agent tool called Eigent
Eigent is a multi-agent collaboration tool developed by the CAMEL-AI team, built on the OWL framework, aiming to improve the efficiency of handling complex tasks. Its core lies in task decomposition and parallel execution, supporting multiple large language models and multimodal data processing, demonstrating the advantages of open-source AI in performance and accessibility.
【AiBase Summary:】
🧠 Eigent achieves efficient task processing through multi-agent collaboration, significantly improving the execution efficiency of complex tasks.
📊 Supports multiple large language models and multimodal data processing, generating more professional and comprehensive content.
🌐 Open-source characteristics allow developers to freely check code, contribute features, or customize usage, promoting community development.
Details link: https://github.com/eigent-ai/eigent
12. 2025 Top 100 Apps with Fastest User Growth Released, DeepSeek Ranked First
The article analyzed the top 20 applications with the highest growth rate in active users exceeding 100 million in June 2025, where DeepSeek stood out in the AIGC field, and Dabao and Hongguo Free Short Drama also showed strong growth momentum.
【AiBase Summary:】
🌟 DeepSeek ranks first in the AIGC industry with 163 million monthly active users.
📈 Dabao achieved a 410.69% year-on-year growth, with 141 million monthly active users.
🎥 Hongguo Free Short Drama attracted 212 million users, meeting the demand for fragmented entertainment.