Welcome to the "AI Daily" section! This 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 technical trends and learn about innovative AI product applications.

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

1. Qwen3-VL Twin Models Open Sourced: A New Paradigm in Multimodal Retrieval

The Tongyi Lab of Alibaba has open-sourced the Qwen3-VL-Embedding and Qwen3-VL-Reranker models, marking a new era of semantic alignment in multimodal information retrieval. These two models achieve precise matching and efficient retrieval of cross-modal content through an efficient dual-tower architecture and single-tower cross-attention mechanism.

image.png

AiBase Highlights:

🚀 Qwen3-VL-Embedding adopts a dual-tower architecture to achieve unified vector representation for heterogeneous content such as text, images, and videos.

🔍 Qwen3-VL-Reranker improves the accuracy and relevance of search results through a cross-attention mechanism.

🌐 Supports over 30 languages, offers flexible vector dimension selection, and high-performance quantized versions, lowering the integration barrier for developers.

Details: https://github.com/QwenLM/Qwen3-VL-Embedding

2. Tencent Tries a New Entertainment Avenue for Gen Z: “Shangtou Frog” AI Interactive Story Mini Program Goes into Trial, Using Generative AI to Stimulate Young Creativity

The article explores Tencent's AI interactive story mini program "Shangtou Frog," which focuses on using AI technology to enable user participation and co-created content, meeting the demand of Gen Z for interactivity and social interaction, while offering new ideas for Tencent's application in the AIGC field.

image.png

AiBase Highlights:

🎮 An AI-driven interactive story experience where users become the master of the plot.

👥 Socially designed to create a user-generated content community.

🚀 Tencent is exploring new forms of entertainment in the generative AI space.

3. XPeng Unveils “Physical World Large Model”: Second-Generation VLA Targets L4, Marking the Beginning of the No-Navigation Smart Driving Era

XPeng released its second-generation VLA large model at the 2026 global product launch event, marking a step forward toward initial L4-level capabilities in intelligent driving systems. The model has the ability to understand, infer, and generate actions, actively simulate traffic dynamics, and predict potential conflicts, while supporting no-navigation automatic auxiliary driving, greatly expanding the coverage of high-level assistance driving.

image.png

AiBase Highlights:

🧠 The second-generation VLA large model achieves a paradigm shift from "perception-decision" to "understanding-inference-generation."

🚗 Supports no-navigation automatic auxiliary driving, covering complex urban side roads and unmarked road areas.

🚀 Cross-domain driving capabilities can be transferred to SUVs, coupes, and future flying cars, building a unified intelligent mobility ecosystem.

4. MiniMax Surges 42% on First Day! Lands on Hong Kong Stock Exchange, Sets Fastest IPO Record, Ignites AI Investment Frenzy

MiniMax, a leading domestic large model company, successfully listed on the main board of the Hong Kong Stock Exchange, with its stock price surging 42.67% on the first day, showing the high recognition of the capital market. Its IPO process set a record for the Hong Kong new share market, becoming the fastest AI company from establishment to IPO globally, marking the beginning of a new era for Chinese large model companies entering the secondary market.

image.png

AiBase Highlights:

🚀 MiniMax listed on the Hong Kong Stock Exchange, with a 42.67% surge on the first day.

💼 Became the fastest AI company from establishment to IPO globally.

📈 2026 marks the beginning of a new era for Chinese large model companies entering the secondary market collectively.

5. GaoDe FantasyWorld Tops World Model Rankings, Another Victory for Alibaba's Spatial Intelligence!

GaoDe Map's self-developed world model FantasyWorld topped the WorldScore Leaderboard, showcasing its technological strength in 3D world construction. The model significantly improves the visual realism and multi-view consistency of generated videos and supports downstream tasks without additional optimization. FantasyWorld has been applied to GaoDe Map's "Flight Street View" feature, providing merchants with high-fidelity 3D virtual street views, reducing the barriers to professional 3D modeling, and promoting developments in embodied intelligence and autonomous driving.

image.png

AiBase Highlights:

✨ FantasyWorld secured the top spot on the WorldScore Leaderboard with its innovative joint modeling approach.

🚀 The model significantly enhances the visual realism and multi-view consistency of generated videos, supporting downstream tasks without additional optimization.

🧭 GaoDe Map has applied FantasyWorld to its "Flight Street View" feature, helping merchants provide high-fidelity 3D virtual street views.

6. One Company Beats Eight! Alibaba Qwen Downloads Lead Globally, Beating Meta and OpenAI Combined in a Month

Alibaba Qwen's large model leads the global open-source ecosystem with a significant advantage, with downloads exceeding 700 million, beating the combined total of global giants like Meta and OpenAI in a single month. Its full-stack open-source strategy enables it to compete with top closed-source models in multiple performance indicators, marking a historic peak in the international developer community for domestic open-source large models.

image.png

AiBase Highlights:

🚀 The Qwen series model has exceeded 700 million downloads, becoming the most widely adopted open-source model by global developers.

🔥 Monthly downloads exceed the combined total of global giants like Meta and OpenAI, showcasing explosive growth.

🌟 Full-stack open-source strategy covers multimodal fields, with performance comparable to top closed-source models like GPT-5-high.

7. Google Gmail Gets Major Update: AI-Powered Inbox and Natural Language Search Now Available

Google Gmail receives a major update, introducing an AI-powered inbox and natural language search features to improve user email processing efficiency.

image.png

AiBase Highlights:

📧 AI Inbox: Automatically organize pending tasks and summarize important updates.

🔍 Natural Language Search: Find email details by asking direct questions.

🎁 Multiple features are available free to all users, lowering the barrier to use AI tools.

8. Musk’s xAI Enters “Vibe Coding,” New Product Grok Build Exposed

Musk's xAI company is launching a new product called Grok Build, whose core concept is "vibe coding," aiming to let AI automatically plan and build code through natural language. The product will include web and local proxy versions and support full-cycle automated development tasks. Musk confirmed that Grok Code will enter a period of intensive updates, signaling a more competitive landscape in the AI programming tool market.

image.png

AiBase Highlights:

💻 New Programming Paradigm: xAI introduces the "vibe coding" solution Grok Build, aiming to let AI automatically plan and build code through natural language.

🛠️ Multi-Form Deployment: The product will include a web-based interface and a local proxy version with CLI, supporting full-cycle automated development tasks.

🚀 Strategic Upgrades: Musk confirmed that Grok Code will enter an intensive update phase, indicating a more competitive landscape in the AI programming tool market.