aimode.news
Published on

AI daily: button 3.0 officially online; bean buns expected to go online in late June; Krea 2 Lora. Full Open

Authors

Welcome to the "Ai Daily" section. This is your daily guide for exploring artificially intelligent worlds, and we give you the hot spots in the field of AI every day, focusing on developers, and helping you to understand technological trends and innovative AI applications.

Fresh AI products click on: https://app.aibase.com/zh

One, the bean bag is expected to go online in late June: third-class subscription prices are exposed and the connection is to be diverted in three quarters

The Bean Pack plans to formally go online in late June and to update the functionality simultaneously at the Force Congress, marking substantial progress towards commercialization of the largest model of domestic users. The bean bag will be used in the third quarter to improve the pay-as-you-go scenario in conjunction with the electrician ' s functionality and to channel the tremor town through targeted subsidies, with a full stabilization of operations in the fourth quarter. In addition, the penetration of paid users will not be used as a core test indicator in the bean bag for the time being in 2026.

[AiBase Summary:]

The bean bag is scheduled to go online in late June to pay for the content and to update the functionality.

In the third quarter, the pay-off scenario will be improved in conjunction with the electrician ' s functionality, leading the tremors through targeted subsidies.

The soybean kit was suspended in 2026 from the core performance indicator for fee-paying user penetration.

2 button 3.0 officially online, supporting multiple Agent co-work

The byte button version 3.0 achieved a significant breakthrough in team collaboration and ecological access, not only supporting a flexible Agent combination model, but also enhancing cross-end synchronization and local ecological integration, while providing customized solutions for multiple vertical industries.

[AiBase Summary:]

The new version supports the "One + More Agent" and "Multi People + More Agent" flexible combination models.

• A seamless transition from the mainstream environment to support multiple local Agent access.

• Provides a fine template for multi-industry core scenarios and helps users quickly build expert-level intelligence.

3, Krea 2 Lora. All open! AI, the era of personalization of image creation is coming.

The full opening of Krea 2 LoRA marks the beginning of an era of personalization of AI image creation, where users can train and use custom styles, roles or objects without high-end subscription, and significantly lower the threshold for AI image creation.

[AiBase Summary:]

The Krea2LoRA training is fully open and users can customize styles, roles or objects.

The Krea2 model, with its high-performance image generation engine at its core, supports rapid response to vague tips and produces high-quality images.

🚀 Krea2LoRA significantly outperformed other models in terms of accuracy, quality and creative performance to enhance the individualization of AI image creation.

4.OpenAISubstantial adjustmentChatGPTSupport the early lockdown of the algorithm.

OpenAI versus ChatGPTSystemic upgrades, including hand-porting interactive logic, long-dialogue catalogue functionality and bottom model optimization, have enhanced the efficiency of user experience and the implementation of productivity tools.

[AiBase Summary:]

High-level gestures are introduced to allow users to choose different response slots to accommodate different tasks.

Launch long-dialogue directory navigation, which allows users to jump quickly to specific topics for discussion.

OptimizationGPT-55 Instant model to improve text modulation and readability and reduce mechanized response.

Five, programming and GUI two-way food!

Ali officially released a new multi-modular large model Qwen3.7-Plus, which fully upgrades visual-linguistic abilities based on text capabilities and unites them into integrated intelligent base. Qwen3.7-Plus allows seamless integration of GUI with CLI and automating the front-end prototype to the end of a complex software project. In the list of authoritative visual models, Ali has succeeded in being among the top five in China.

[AiBase Summary:]

• Pure text and reasoning Agent: Strong performance on complex software engineering and scientific programming tasks.

🖼️ Multi-modular reasoning and visual programming: Supports the conversion of the first key of images, videos, UI screenshots into enforceable codes.

• Real world perception and video understanding: covering document resolution, advanced OCR, long video event stream understanding.

6 . Micro-Intelligence Exposure: Main Call Inner Task Processing, phased up on-line

A prototype of an AI smart body is being tested to help users perform a number of tasks efficiently within the Web. The roll-out of the smart body would activate the ecological potential of Twitter as a SuperApp and could recreate the service connectivity mechanism within it.

[AiBase Summary:]

📱 Micro-Screen is testing the prototype of the AI smart body to improve the efficiency of user task handling.

🔄 The smart body will go online in stages, starting with a closed test and then gradually expanding its scope.

🌐 This placard will fully activate the ecological potential of Micro-Intelligence and facilitate the transition of the mobile Internet to AI Agent, the portal.

7 Subversive Traditions! OpenAI's live demonstration of "No App" phones, all interfaces being generated in real time by AI

The article presents the prototype of the “Agenda Operating System” presented at the event by OpenAI, which produces real-time interfaces through voice commands without the need to install third-party applications, and recalculation missions are handled by cloud-based GPT megamodels, which demonstrate subversive changes in the way smartphones interact in the future.

[AiBase Summary:]

• The system produces real-time interfaces through voice commands without the need to install any third-party applications.

• Recalculative tasks are handled by cloud-based GPT large models, which achieve efficient operation.

OpenAI plans to produce AIS smartphones in the first half of 2027 to facilitate the process of industrialization.

8.GitHub Copilot officially goes on-line in a volume billing mode

GitHub Copilot's official up-line billing model, introducing the AI point deduction mechanism, marks its transition from a single code completion tool to an intelligent platform.

[AiBase Summary:]

💡 GitHub Copilot formally uses the usage-based model, which requires users to consume AI points to cover service costs.

• The model changed the old low-priced infinity package month to shift costs to the user-end.

• High-value missions such as code review and cloud-based intelligent body missions will be factored into point consumption, affecting the heavy user budget.

AI daily: button 3.0 officially online; bean buns expected to go online in late June; Krea 2 Lora. Full Open | aimode.news