AI Daily: KlingAIAvatar 2.0 Launches; Google Introduces Gemini 3 Deep Think Mode; Alibaba Cloud XiYan-SQL Strongly Wins

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, helping you understand technology trends and innovative AI product applications.

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

1. KlingAI Avatar 2.0 Launches and Goes Viral: Generate a 5-Minute Dance and Sing in One Click, Digital Humans Officially Say Goodbye to the "Stiff Face" Era

KlingAI Avatar 2.0 achieves intelligent transformation from audio to emotional performance through a multimodal director module, significantly enhancing the expressiveness and movement of digital humans, bringing revolutionary impact to fields such as short videos, e-commerce ads, and educational content.

AiBase Summary:

✨ Avatar 2.0 converts audio, images, and text prompts into a coherent storyline through its multimodal director module.

💡 Achieves a qualitative leap in facial expression control and action design, avoiding the "stiff face" feel of early AI characters.

🚀 Supports ultra-high frame rates of 48fps and 1080p high-definition output; users can try basic functions for free on the platform.

Details: https://app.klingai.com/cn/ai-human/image/new

2. Google Launches Gemini 3 Deep Think Mode, Significantly Enhancing AI Reasoning Capabilities

Google has launched the Gemini 3 Deep Think mode, significantly improving AI's reasoning capabilities, especially in handling complex mathematical, scientific, and logical problems. The mode performs exceptionally well in multiple benchmark tests, achieving a score of 41.0% in the "Final Exam of Humanity," and reaching 45.1% in the ARC-AGI-2 test when using code execution. This improvement is due to its advanced parallel reasoning technology, which can explore multiple hypotheses simultaneously. Ultra subscription users can easily experience this feature, further advancing AI technology.

AiBase Summary:

🧠 Gemini 3 Deep Think mode officially launched, enhancing reasoning capabilities and focusing on complex problems.

📊 Demonstrated excellent performance in strict benchmark tests, scoring 41.0% without tools and 45.1% with code use.

🚀 Ultra subscription users can experience this powerful mode with simple selection, promoting the advancement of AI technology.

Details: https://blog.google/products/gemini/gemini-3-deep-think/

3. Doubao Mobile Assistant Announces Adjustments: AI Operation Capabilities Enter a Standardized Phase

Doubao Mobile Assistant released an announcement stating that it will standardize some AI operation capabilities of mobile phones to maintain the platform ecosystem and financial security.

AiBase Summary:

📱 AI phone operation features require user authorization; they can be terminated at any time during execution.

🔒 Restricts AI from performing automated operations such as score brushing or incentive brushing within apps.

💰 Further restricts proxy operations in financial applications like banks and internet payments.

4. Microsoft Releases VibeVoice 0.5B: Achieving Real-Time Speech in 300 Milliseconds with Only 0.5B Parameters

Microsoft released a new real-time text-to-speech model called VibeVoice-Realtime-0.5B, offering new possibilities for AI voice interaction with its compact size and strong performance. The model supports real-time transcription and speech generation in both Chinese and English, maintains unique tones, rhythms, and voices in multi-character dialogues, and possesses emotional expression and context memory capabilities, making the speech more natural and realistic.

AiBase Summary:

🧠 Small model size but strong performance, achieving near real-time speech generation with only 0.5B parameters.

🗣️ Supports real-time transcription and speech generation in both Chinese and English, naturally presenting multi-character dialogue scenarios.

💡 Possesses emotional expression and context memory capabilities, making speech closer to human expression.

Details: https://huggingface.co/microsoft/VibeVoice-Realtime-0.5B

5. Preview of the Android XR Special Event: Gemini-Powered Smart Glasses Make Their Debut, Can Google Regain Ground with "Spatial Computing"?

The article introduces Google's upcoming special event "The Android Show: XR Edition," highlighting software updates and hardware ecosystems of the Android XR platform, including software base upgrades, hardware ecosystem presentations, and content for developers and usability.

AiBase Summary:

🧠 Software base upgrade, optimizing system response speed and multi-device collaboration, with simultaneous updates to third-party developer toolchains, reducing adaptation costs for hardware manufacturers.

👓 Hardware ecosystem presentation, Samsung Galaxy XR headset will be demonstrated, and smart glasses prototypes may be publicly revealed for the first time.

🚀 Developers and usability, technical documentation and replays will be open after the live stream, and the Android XR SDK will include Gemini Runtime.

6. The Strongest Coding Model Launched! GPT-5.1-CodexMax Now Integrated with Response API

OpenAI announced that its latest and most powerful agent coding model, GPT-5.1-CodexMax, has been fully integrated with the response API, allowing developers to directly integrate this top-tier coding intelligence into existing applications and production workflows. The model shows significant improvements in complex task decomposition, code generation quality, multi-step reasoning, and autonomous agent execution. With the official release of the API, developers can call this flagship model in a broader environment without waiting. Users who call CodexCLI with API keys have also gained access to GPT-5.1-CodexMax. OpenAI stated that this update aims to further lower the entry barrier for high-performance AI programming capabilities, enabling more products and services to enjoy the experience of "always ready to write, automatically correct, and self-execute" programming assistants.

AiBase Summary:

🧠 GPT-5.1-CodexMax is OpenAI's latest powerful coding model, capable of improving complex task decomposition and code generation quality.

🚀 The model has been fully integrated with the response API, allowing developers to directly integrate it into existing applications and workflows.

💡 OpenAI stated that this update aims to lower the entry barrier for high-performance AI programming capabilities, enabling more products and services to enjoy the programming assistant experience.

7. Aliyun Xiyansql Dominates Globally, Wins the SQL Diagnostic Evaluation List!

The data analysis intelligent entity "Xiyansql" developed by Alibaba Cloud's Feitian Lab performed outstandingly in the BIRD-CRITIC evaluation, successfully topping all open lists, surpassing many top teams both domestically and internationally, setting a new industry record for SQL diagnosis and repair. The evaluation covered major database systems such as MySQL, PostgreSQL, SQL Server, and Oracle, with questions ranging from simple queries to complex operations, with overall difficulty far exceeding traditional tests. Xiyansql improved the model's executability and maintainability through innovative methods and has now been launched on Alibaba Cloud's BaiLian platform, providing SQL generation and diagnostic services.

AiBase Summary:

✅ Xiyansql won first place in the BIRD-CRITIC evaluation, surpassing many top teams.

📊 The evaluation covers multiple major databases, with higher difficulty than traditional SQL generation tests.

💻 Related technologies and models have been open-sourced, supporting developers for experience and contribution.

8. OpenAI Launches GPT-5.1-Codex-Max

OpenAI's GPT-5.1-Codex-Max excels in both performance and price, with strong coding capabilities and optimization for Windows environments, making it important in the developer market.

AiBase Summary:

🧠 GPT-5.1-Codex-Max shows significant performance improvements, with pricing consistent with GPT-5.

⚙️ The model has "agent-style" coding capabilities and long-running characteristics, capable of handling complex tasks.

💻 OpenAI optimized for Windows environments and integrated into multiple development tool ecosystems.