AI Daily: Ant Open Sources Large Model Ming-flash-omni 2.0; Zhipu's GLM-5 Leaked Unexpectedly; JD.com Officially Enters the AI Payment Field

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 technology trends and innovative AI product applications.

Discover new AI products click to learn more:https://app.aibase.com/zh

1. Ant Group Open Sources the Multimodal Large Model Ming-flash-omni 2.0: Significant Improvements in Multimodal Understanding, Image Editing, and Speech Generation

Ant Group has open-sourced the multimodal large model Ming-flash-omni 2.0, which performs exceptionally well in multiple public benchmark tests, setting a new performance standard for open-source multimodal large models. The model achieves leading levels in core areas such as visual language understanding, speech controllable generation, image generation and editing, and has unified audio generation capabilities, providing a unified entry point for multimodal application development.

【AiBase Summary:】

🧠 Ming-flash-omni 2.0 excels in key capabilities such as visual language understanding, speech controllable generation, and image generation and editing, with some metrics surpassing Gemini 2.5 Pro.

🔊 As the industry's first full-scenario audio unified generation model, Ming-flash-omni 2.0 can generate speech, ambient sound effects, and music simultaneously on the same audio track, supporting fine-grained parameter control via natural language.

🚀 It maintains a leading edge in reasoning efficiency and cost control, achieving real-time high-fidelity generation of long audio within minutes, significantly improving the efficiency of multimodal application development.

2. Zhipu GLM-5 "Leaked" by Accident? Reusing DeepSeek Architecture, Performance Blows Up, Market Cap Soars 200%

Zhipu AI's next-generation large model GLM-5 gained attention during the 2026 Spring Festival, with its performance and technological innovation receiving high market recognition, causing its stock price to surge 200%.

【AiBase Summary:】

🧠 GLM-5 uses the same sparse attention architecture (DSA) as DeepSeek-V3, significantly enhancing performance.

🧮 The parameter count reaches 745B, twice that of the previous generation GLM-4.7, with high computational efficiency.

🎥 Enhances multi-modal capabilities such as video understanding, overcoming the shortcomings of DeepSeek's pure text architecture.

3. No Longer Dependent on Others for Computing Power! iFLYTEK Officially Launches Spark X2 Large Model: Nationally Developed Computing Power Training, Focused on Four Professional Scenarios

iFLYTEK launched the Spark X2 large model, trained using nationally developed computing power, focused on four professional scenarios, promoting the process of domestic replacement, and providing safer and more reliable technical support for the industry.

【AiBase Summary:】

🧠 The Spark X2 large model is trained using domestically developed computing power, achieving self-control and autonomy.

📚 Focuses on four professional scenarios: education, healthcare, automotive, and intelligent agents, enhancing industry application value.

🔒 The deepening of domestic substitution provides a safer and more reliable large model option for key industries in China.

4. JD.com Enters the AI Payment Field! Launches "JD AI Pay": Speak and Buy, Enhanced Payment Security

JD Technology officially launched the innovative payment product "JD AI Pay," based on the JoyAI large model, allowing users to complete payments through voice commands, improving convenience and security. Additionally, the product is applied on smart glasses and other devices, demonstrating significant scenario potential.

【AiBase Summary:】

🤖 Voice as Payment, Achieving "See and Buy Immediately"

🔒 Dual Verification Technology, Protecting Fund Security

📈 Industry Insight: Major Players Compete in the New AI Payment Market

5. Free to Use: DuckDuckGo AI Voice Chat Launched, Commitment to Not Store Audio

DuckDuckGo added a real-time voice chat feature through its AI chatbot platform, Duck.ai, emphasizing privacy protection. This feature adopts a "privacy-first" architectural design, allowing users to have natural conversations with large language models through encrypted channels, ensuring that voice data is not monitored or reused. DuckDuckGo establishes a firewall between users and OpenAI, with both parties under strict contract limitations to ensure data security. In addition, Duck.ai offers both registration-free use and paid subscription services to meet different user needs.

【AiBase Summary:】

🔒 Privacy Commitment: The official clearly states that no chat audio will be stored, and all session data will be destroyed immediately after the conversation ends.

🛡️ Zero Data Training: Users' voice and AI responses will not be used for any algorithm model training.

🆓 Flexible Usage: Supports direct use without registration for voice functions, and also provides paid subscriptions to meet higher frequency commercial or personal needs.

6. Intelligent Driving and Cabin Evolve Together! AVATR.OS 5.0 of Avita Officially Released: MoLA Large Model Onboard, First Arriving at Huawei ADS 4.1

Avita officially released the AVATR.OS 5.0 system, integrating AI large models and the Huawei ADS 4.1 intelligent driving system, bringing smarter in-car assistants, safer driving assistance, and more personalized cabin experiences.

【AiBase Summary:】

🧠 MoLA large model enhances semantic understanding and interactive experience.

🚗 Huawei ADS 4.1 intelligent driving system optimizes urban navigation and safety defense.

📱 HarmonyOS cabin enables personalized customization and convenient connectivity.

7. The Sky Is Falling for Car Insurance Brokers! ChatGPT Launches "Price Comparison Tool": Powered by 190 Million Data Points, the Era of Transparent Premiums Has Arrived

The article discusses the impact of AI technology on the traditional insurance brokerage industry, particularly how Insurify's ChatGPT-exclusive insurance price comparison app changes the car insurance purchasing experience.

【AiBase Summary:】

🚗 Personalized Premium Estimation: Users input location, vehicle type, etc., and ChatGPT instantly generates customized premium estimates.

📊 Massive Data Support: The app accesses a database of 196 million auto insurance quotes and 70,000 customer reviews.

💬 Conversational Shopping Experience: Users can ask questions in simple language, and the system compares prices, discounts, and other key information.

8. GPT-5.2 Drives Strongly! OpenAI Upgrades Deep Research Tools, Unlocking a New Full-Screen Report Interaction Experience

OpenAI has made a major update to the deep research tools of ChatGPT, introducing the core engine driven by the GPT-5.2 model, and adding a full-screen viewer function to improve the processing efficiency and readability of complex information. Users can now specify data sources to be fetched and intervene in real time during report generation, while also supporting downloads in various formats.

【AiBase Summary:】

🖥️ Full-Screen Interaction Upgrade: Added a full-screen viewer that supports directory navigation and source reference comparison, greatly optimizing the reading experience of long in-depth reports.

⚙️ Model Engine Replacement: The deep research tool has now fully integrated the GPT-5.2 model, providing more accurate logical reasoning and global information retrieval capabilities.

🔍 Deep Customization Permissions: Supports specifying data source fetching and allows users to monitor progress and dynamically adjust research scope during report generation.