OpenAI unlocks Deep Research: Paying users can query 10 times per month, and Microsoft releases multimodal AI agent Magma

ChatGPT developer OpenAI announced the launch of a new AI agent feature, Deep Research, for Pro users at the beginning of the month. On the 26th, it was announced that it is now fully open to all paying users. Meanwhile, Microsoft today Open Sources the multimodal AI agent base model, Magma, which can handle multimodal data such as text, images, and videos. (Background: OpenAI counters Musk's malicious acquisition! Proposing 'special voting rights' to the non-profit board to prevent malicious acquisition) (Context: Musk bids 97.4 billion pounds for acquisition of OpenAI! Sam Altman instantly rejects and retorts: Then I'll buy X (Twitter) for 9.74 billion pounds) Chinese AI startup DeepSeek initiated an 'Open Source Week' event this week, gradually Open Sourcing five code repositories in a completely transparent manner to share research progress. At this juncture, competitors such as OpenAI are actively enhancing their technical capabilities by not only introducing enhanced features to improve model performance but also accelerating Open Source initiatives in certain areas. OpenAI announced the launch of the new AI agent feature, Deep Research, for Pro users at the beginning of the month, opening it up for Pro users. This feature can carry out multi-step research online, designed for complex tasks, and can complete tasks that originally required hours of manual work in just a few tens of minutes. Users only need to provide a prompt, and ChatGPT will search, analyze, and synthesize hundreds of online sources to produce a professionally crafted full report. This feature is powered by OpenAI's o3 model, optimized for web browsing and data analysis, utilizing reasoning to search, interpret, and analyze a large amount of online text, images, and PDFs, and adjust research direction flexibly based on new findings. OpenAI introduced the Deep Research feature to Plus users and now on the 26th, further announced that the Deep Research feature is now fully open to ChatGPT Plus, Team, Edu, and Enterprise users. Some improvements have been made to the Deep Research feature since its initial launch: Embedded images with citations in the output, better understanding and referencing of uploaded files. Plus, Team, Enterprise, and Edu users will be able to use the Deep Research feature 10 times per month, while Pro users will have 120 uses per month. OpenAI also released system cards detailing the development, capability assessment, and security improvements of the Deep Research feature, inviting experts to participate in training future models. Microsoft Open Sources the multimodal AI agent base model, Magma, today on its official website. Compared to traditional agents, Magma has multimodal capabilities across digital and physical worlds, automatically processing different types of data such as images, videos, and text. Additionally, Magma includes built-in psychological prediction capabilities, enhancing its understanding of future temporal and spatial variations, accurately predicting the actions and intentions of people or objects in videos. Users can utilize Magma to perform various automated tasks, such as online shopping, weather inquiries, or even remotely controlling physical robots, providing decision-making suggestions in real chess games to assist users in playing more effectively. According to Microsoft's official introduction, Magma can help AI-driven assistants or robots understand their surroundings and take appropriate actions, enabling home robots to learn how to organize previously untouched items or assisting virtual assistants in generating detailed step-by-step instructions for unfamiliar tasks. As a Visual-Language-Action (VLA) based model, Magma can adapt to new challenges in digital and physical environments, learning from vast public visual and language data. By integrating language understanding, spatial perception, and temporal reasoning capabilities, the model can handle various complex scenarios, delivering significant intelligent application value in both virtual and real-world environments. Related Reports: OpenAI's first self-developed AI chip is expected to be designed and handed over to TSMC for trial production this year to balance NVIDIA's dominant position? Altman's Law of Sam: The cost of using AI will decrease by 10 times every year, making AI as cheap as air in the future. OpenAI freely opens the 'ChatGPT search function', no need to register for an account, making Google anxious? 'OpenAI unlocks Deep Research: paying users can query 10 times a month, and Microsoft releases the multimodal AI agent Magma.' This article was first published on BlockTempo, the most influential blockchain news media in the Block sector.