月之暗面在Hugging Face上线Kimi K3权重开源倒计时
Related
Baidu's OCR model, Unlimited OCR, topped four charts: HuggingFace, GitHub, and more.
According to Mars Finance, Baidu recently officially released and open-sourced its end-to-end OCR model, Unlimited OCR. The model topped the GitHub Daily Trending and Python charts the day after its release, and also ranked first on HuggingFace's global overall model trending chart and multimodal model trending chart. Unlimited OCR is designed for long document parsing scenarios, with a total parameter size of 3B and only about 570M activation parameters during inference. Public evaluation results show that Unlimited OCR achieved a 93.92% overall score in the OmniDocBench v1.6 benchmark test, setting a new record for end-to-end OCR.
AI-powered agents suffer setbacks in their first foray into coffee shops: Gemini's excessive discounting leads to losses, and GPT's overly stingy practices cause raw material shortages.
According to Beating's monitoring, AI evaluation agency Andon Labs released test data on its AI agent Mona operating a physical coffee shop. In the first two months, Mona ran on a Gemini 3.1 Pro model. During this period, the model showed almost no concept of profit, not only excessively purchasing raw materials but also being easily swayed by customer claims, offering large discounts or even free items, and even admitting to a customer's claim of a 99% discount without verification. This resulted in the coffee shop spending approximately $15,000 on supplier and equipment purchases, while sales were only $9,000, leading to a net operating loss of nearly $6,000 (if fixed costs such as rent and salaries are included, total expenditures reach $38,000). Subsequently, the team switched the model to GPT-5.5. The new model showed significant anxiety in the face of losses and immediately stopped blindly ordering. However, this went to the other extreme: insufficient purchases led to a shortage of fresh raw materials. As of June 25, the availability of menu items had dropped to 77%, and 10 dishes had been forced to be removed from the menu. Meanwhile, GPT-5.5 demonstrated extremely strong anti-cheating and anti-jailbreak capabilities, rejecting all customers who requested special prices or offered free food in exchange for social media promotion.
Google Antigravity version 2.2.1: Integrates system key ring and fixes multi-agent deadlock.
According to Beating's monitoring, Google Antigravity has released version 2.2.1. Addressing the frequent issue of authorization expiration faced by developers, the new version supports automatically writing refreshed OAuth credentials into the operating system's security keyring (such as Apple Keychain or Windows Credential Manager), eliminating the need for repeated web page authorization pop-ups. Simultaneously, workspace search has been upgraded to substring matching mode, fixing path non-existence errors caused by prefix search and resolving deadlock issues in sub-agent operation. Multimedia and environment compatibility have also been optimized. The new version adds direct playback capabilities for .mp3 and .wav audio formats in the sidebar and deliverables window, and adds C++, Python, and Protobuf syntax highlighting to Markdown. For Windows platform users, the update fixes repeated triggering of UAC access control and errors accessing the system environment variable PATH directory. The system also includes a built-in Antigravity Guide to answer platform usage questions. However, some developers have encountered obstacles during the upgrade and experience. Community feedback indicates that an issue with the upgrade channel has caused some users to be stuck on the older 2.1.4 version with no available updates. Additionally, the client is forced to default to a resolution of 1400x900 pixels each time it launches, and the new version has removed the inline command editing functionality previously triggered by Ctrl+I, drawing numerous user complaints.
Notion is launching a public beta test of the Claude and Cursor agents: Kanban-based direct task assignment and support for external subscriptions.
According to Beating, Notion announced a public beta of its External Agents platform in its Business and Enterprise programs, supporting the integration of Anthropic's Claude and AI programming tool Cursor into team collaboration dashboards. Teams can @ agents within the Notion workspace, similar to mentioning colleagues, or directly assign tasks on task boards, allowing AI to automatically collaborate in the background to complete tasks such as code writing, data analysis, and content generation. For general office users, the newly beta Claude agent runs on the Anthropic Managed Agent (CMA) architecture. Users do not need to configure complex API keys and can pay directly using Notion credits per run. After assigning task cards to agents, the model can autonomously read related documents and execute multi-step planning. However, it should be noted that because the CMA architecture operates in a stateful manner, Notion's existing Zero Data Residue (ZDR) commitment does not apply to Claude agents; session data will remain on the server side. For development teams, Notion has opened up an External Agent interface. Users can bind their locally running Cursor agent to a Notion workspace. For example, after assigning a coding task on a Notion Kanban board, Cursor will use the user's personal subscription to complete the code development and submit a Pull Request in the local runtime environment. Finally, it will automatically write the execution results and PR link back to the Notion task page, achieving development automation with zero Notion points consumption.
Qwen-AgentWorld published by Qianwen
Mars Finance reported on June 24th that Qianwen officially released its native language world model, Qwen-AgentWorld, capable of simulating intelligent agent interaction environments across seven domains. Also released simultaneously was AgentWorldBench, a benchmark covering all seven domains. The model and benchmark are available from Hugging Face and ModelScope. (Wide Angle Observation)
Google Gemini launches Interactions API: Reconstructing intelligent agents to end "phantom bills".
According to Beating's monitoring, Google has officially made the Interactions API the default interface in the Gemini API and Google AI Studio, while the original interface generateContent has entered maintenance. The new interface will package multi-turn dialogues, model thinking, tool calls, and execution results into a unified package and store them as Interaction resources on the server side. The most direct benefit is cost-effectiveness and labor-saving. Since the server automatically saves the dialogue state, during multi-turn chats, the client only needs to send the interaction ID of the previous step, and Google can automatically reuse it and improve the cache hit rate, significantly reducing token fees. For time-consuming tasks such as code sandbox running or deep search, developers can enable background mode to allow the work to run asynchronously in the background. In addition, intermediate steps such as model thinking and tool calls are fully visible, making it easy to troubleshoot bugs in the operation of intelligent agents in real time. The underlying refactoring directly solved the billing crisis that broke out in early June. At that time, after manually deleting the cache, due to cloud synchronization delays, the cancelled resources were still running in the background, resulting in some accounts being abnormally deducted nearly 27,000 RMB within 12 hours. Additionally, the model is prone to getting stuck in a thought loop during online searches, still charging the full amount of 64,000 tokens even in a "zero output" state with no responses. The Interactions API replaces manual deletion with a server-hosted "implicit cache" and allows developers to monitor and halt the thought process, providing a more secure billing defense. Since Google's future cutting-edge models and agent features (such as Deep Research) will only support the Interactions API, developers are officially advised to complete the adaptation as soon as possible.