Welcome to the "AI Daily" section! Here 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 to help you understand technology trends and innovative AI product applications.

New AI products Click to learn more:https://app.aibase.com/zh

1. ByteDance launches Dou Bao Large Model 1.6: The first domestic model supporting adjustable thinking depth

ByteDance's Volcano Engine launched Dou Bao Large Model 1.6, which for the first time supports adjustable thinking depth, improving the balance between efficiency and quality, and introducing a lightweight version to meet enterprise needs.

image.png

【AiBase Highlights:】

🧠 Dou Bao Large Model 1.6 supports adjustable thinking length, improving the balance between efficiency and quality.

💼 The Dou Bao 1.6lite version optimizes enterprise scenarios, reducing usage costs.

📈 Tiered mechanisms solve the problem of resource waste in traditional models, aligning closely with practical needs.

2. Baidu Launches the World-Leading Document Parsing Model PaddleOCR-VL, Reshaping the OCR Technology Landscape!

Baidu's PaddleOCR-VL model has shown excellent performance in document parsing. With its lightweight and efficient features and outstanding performance, it has achieved excellent results in multiple evaluations. The model supports multiple languages and is applicable to various intelligent document processing tasks.

image.png

【AiBase Highlights:】

✨ PaddleOCR-VL ranks first globally in OmniBenchDoc V1.5 with 92.6 points, demonstrating core capabilities such as text, tables, and formulas.

🔍 The model has only 0.9B parameters and supports 109 languages, suitable for government and enterprise document management, knowledge retrieval, and other scenarios.

🚀 The inference speed has significantly improved, processing 1881 Tokens per second, showing a clear advantage over other mainstream models.

3. AiShi Technology Completes a 100 Million RMB B+ Round Financing: ARR Exceeds 40 Million USD, Users Exceed 100 Million

AiShi Technology has made significant progress in the AI video generation field, completing a 100 million RMB B+ round financing, and achieving an ARR breakthrough of 40 million USD and more than 100 million registered users. Its products enhance user engagement through social operations and localized creative preferences, while the open API system has also attracted a large number of third-party developers.

image.png

【AiBase Highlights:】

🚀 AiShi Technology completed a 100 million RMB B+ round financing, indicating market recognition of its technology and business model.

📈 ARR exceeds 40 million USD, with more than 100 million registered users, indicating that its products have broad market appeal.

🌐 After opening its API system, more than 10 million videos were generated, proving that its technical capabilities have been widely validated.

4. Anthropic Launches Claude “skills” Feature to Enhance AI Work Efficiency

Anthropic launched a new feature called 'skills' for the Claude AI chatbot, aiming to improve the practicality of AI agents in work. This feature consists of a series of folders containing instructions, scripts, and resources, enabling Claude to demonstrate stronger capabilities in specific tasks. Users can also create custom skills according to their needs and use these skills across multiple platforms. This feature echoes OpenAI's AgentKit, showing that the AI industry is moving toward more practical directions.

image.png

【AiBase Highlights:】

🛠️ Users can create custom skills to better adapt Claude to specific work scenarios.

🚀 This move coincides with the release of new features like AgentKit by OpenAI, showing the continuous shift of the AI industry towards practicality.

🌟 Anthropic launched the Claude “skills” feature to enhance the practicality of AI in work.

5. Pinterest Launches AI Content Limit Tool: Users Can Customize Reduce Generative AI Images

Pinterest launched a new AI content limit tool, allowing users to customize the display ratio of generative AI images to address user dissatisfaction with AI content overload. This feature allows users to adjust the display of AI content in specific categories and optimize the experience through feedback mechanisms.

image.png

【AiBase Highlights:】

🖼️ Pinterest launched a new content control tool, allowing users to limit the proportion of AI-generated content in their feed.

⚙️ Users can select to reduce AI-generated images in specific categories, such as beauty, art, fashion, and home decoration, in the settings menu.

🔄 While embracing AI technology, Pinterest is trying to protect user experience, balancing human creativity with AI innovation.

6. Fully Open-Source LLaVA-OneVision-1.5, a Multimodal Model Surpassing Qwen2.5-VL, Has Arrived

LLaVA-OneVision-1.5 is an open-source multimodal model capable of handling various inputs such as images and videos, and it performs well in multiple benchmark tests, surpassing the Qwen2.5-VL model.

image.png

【AiBase Highlights:】

🧠 LLaVA-OneVision-1.5 is a new multimodal model that can handle various input formats, including images and videos.

📈 The training process is divided into three stages, aiming to efficiently enhance the model's visual and language understanding capabilities.

🏆 In benchmark tests, LLaVA-OneVision-1.5 performed excellently, surpassing the Qwen2.5-VL model.

Details link: https://github.com/EvolvingLMMs-Lab/LLaVA-OneVision-1.5 https://huggingface.co/lmms-lab/LLaVA-OneVision-1.5-8B-Instruct

7. OpenAI Video Generation Model Sora 2 Goes Live on Microsoft Azure: Pricing at $0.1 Per Second, Enters Public Preview Stage

Microsoft announced that OpenAI's Sora 2 video generation model has been launched on the international version of Azure AI Foundry and has entered the public preview stage. The model supports multimodal input and is suitable for advertising production, educational videos, and other scenarios. The pricing is $0.1 per second, but currently only available to international users.

image.png

【AiBase Highlights:】

🎥 Sora 2 is a video generation model developed by OpenAI, and it is the first time that the API interface is opened to enterprises through Azure AI Foundry.

💰 The pricing is $0.1 per second, suitable for enterprise users who need to generate short videos in bulk.

🌐 Sora 2 is currently only available on the international version of Azure AI Foundry, and Chinese users cannot access it for now.

8. Travel Search Engine Kayak Launches "AI Mode" for More Convenient Travel Planning and Booking

Kayak launched a new "AI Mode," which helps users research, plan, and book travel through an integrated chatbot. This feature uses ChatGPT technology to provide more context-aware search results and supports open-ended questions to get travel recommendations.

image.png

【AiBase Highlights:】

🌍 Kayak launched "AI Mode," allowing users to easily plan and book travel through a chatbot.

🗣️ This feature supports asking for travel advice and comparing various travel services, using ChatGPT technology to provide accurate information.

📅 "AI Mode" initially supports only English, and will later expand to more languages and platforms, adding voice request functionality.