GitHub

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

1. HangingGPT-5Five!Claude. Opus 4.8 on line, two-thirds drop in price.

AnthropicClaude Opus 4.8, published, achieved significant improvements in intelligent programming, complex logic and multidisciplinary knowledge work, while optimizing user experience and costing.

[AiBase Summary:]

• Enhancement of intelligent body programming and complex logic

📈 Beyond multiple benchmark testsGPT-5.5 andGemini.3.1 PRO

Increased speed of operation and reduced cost of calculation

2. Little red book PC endline AI Search Assistant Point: Multiple rounds of dialogue and collection import based on big, real notes

The little red book goes online at the end of the PC-based AI Search Assistant's Point, which provides active interactive dialogue based on big, real notes, supports multi-wheel dialogue and collection import, and enhances the accuracy of answers and individualized experiences. This function further complements the multi-end interface and breaks down barriers to traditional search engines, deepens the AI search strategy layout, and reshapes the efficiency of content community traffic access and information distribution through the " Generating AI+ Real Community Ecology" differential route.

[AiBase Summary:]

🤖 Little Red Book PC Upline AI Search Assistant Point to support multi-wheel dialogue and collection import

• Breaking down barriers to traditional search engines through the `generated AI+real community ecology' differential route

• Deepen the search strategy layout of AI and reshape the efficiency of content community access and information distribution

Step 3.7 Flash, big model.

Step 3.7 Flash, a new generation of Flash models designed for the productive landing of intelligent bodies, was released and opened, with in-depth optimization of code writing, network search and multimodular workflows.

[AiBase Summary:]

• Using an advanced MOE structure with a total of 196 B.

• A strong primary multimodular understanding capability to directly recognize and interpret complex visual information such as UI interfaces, charts and documents.

🔧 Demonstrating a high degree of instrumentality and stability in long-link, multi-wheel intelligent workstreams.

Four, three, three, four, three, four, three, three, four, three, three, four, three, three, four, one, three, four, one, one, three, three, four, three, four, one, three, three, four, three, three, four, three, three, four, three, three, four, three, four, three, three, four, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, four, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, four, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three, three.

Over the recent past, the control of AI-related non-compliance goods and services has been strengthened, especially in the areas of examination, such as the AI grant tool and the focus on the clean-up of hidden disguised services. At the same time, the Platform has strengthened the governance of the AI-generated content label and introduced a number of comprehensive compliance measures to enhance the safety and health of the consumer shopping environment.

[AiBase Summary:]

• More and more on the management of AI-related non-compliant goods and services, with a focus on the fight against AI, data poisoning, etc.

• The Platform strengthens compliance requirements by specifically governing the AI-generated content marker to ensure compliance with the Artificial Intelligence-generated synthetic content identifier approach.

More than 40 comprehensive governance initiatives have been put in place in the first half of the Platform, covering the core elements of commodity access, advertising governance and intellectual property protection.

5, Grok Build 0.2.7 Heavy pounds update: /usage, /login online, multi-sub-agent shared terminal + image understanding greatly enhanced

The upgrade of Grok Build 0.2.7 resulted in a number of core improvements, including the addition of new /usage and /login commands, the optimization of user management and login processes, the introduction of multi-agent shared terminals, a significant improvement in image understanding and inter-platform compatibility, and a more efficient coding experience for developers.

[AiBase Summary:]

/usage commands are online to enhance user management ease.

/login supports optimization and enhances account switching experience.

Multi-subagent shared terminal functionality and enhanced parallel task processing capacity.

Further links: https://x.ai/build/changelog

MiMo-V2-Pro/Omni announced that by June 2026, MiMo-V2-Pro/Omni would be closed and fully transitioned to the V2.5 series

Mi announced that the old MiMo model service would be discontinued by 30 June 2026 and replaced by the MiMo V2.5 series. This is intended to enhance reasoning and value for money while facilitating the migration of developers to the new version.

[AiBase Summary:]

🚀 MiMo-V2-Pro/Omni model is replaced by MiMo V2.5 series by Mi-Mi.

• The new version of the model has greater reasoning and higher value for money.

Mi enhances the efficiency of the developer ' s access and platform stability by harmonizing model versions.

7. Comprehensive CLI and open-source: a line of commands to get through AI Agent 's full capacity

The full-scale CLI and open source of Aliyun, through a line of commands to access the full-scale organization of AI Agent, achieves a strategic shift from a single-point model to a multi-capacity complex and synergistic approach, while lowering the threshold for developers to build complex task streams.

[AiBase Summary:]

🧠 Ariyun's CLI encapsulates the core competencies of mainstream models, applied workflows, etc., into the command line entrance

• A strategic shift from a single-point model to a multi-capacity, complex and synergistic approach Viet.

• Open source initiative lowers the threshold for developers to build complex task streams

8, 45% amnesia, 35% Token!

The Hy-Memory Smart Memory Plug-in, which was launched by the Quest Mixer, addresses the long-term collaborative, AI Agent memory fragmentation and time-series disorder through three layers of core technology. Hy-Memory performed well on multiple test sets, significantly increasing memory density and processing efficiency and providing multiple configurations to suit different needs.

[AiBase Summary:]

Six layers of memory framework is accurate, improving memory efficiency

⚙️System1/System2 Double-System Balance Speed and Depth

♪ The evolutionary chain keeps its causal traces and avoids historical errors