Welcome to the "AI Daily" section! This is your guide to exploring the world of artificial intelligence every day. Each day, we present you with the latest content in the AI field, focusing on developers, helping you understand technical trends and innovative AI product applications.
Fresh AI products Click to learn more:https://app.aibase.com/zh
1、An AI assistant that generates applications in 30 seconds has arrived! Ant Group's Lingguang App is officially launched
Ant Group's all-modal general AI assistant 'Lingguang' can generate small applications in natural language within 30 seconds, and supports editing, interaction, and sharing. Its three functions include 'Lingguang Dialogue', 'Lingguang Flash Application', and 'Lingguang Eye', which can output full-modal information such as 3D, audio-visual, and charts, greatly improving users' information acquisition efficiency and experience.
【AiBase Summary:】
✨ Lingguang realizes the generation of small applications in 30 seconds through natural language, enhancing user productivity.
🎨 Lingguang Dialogue supports structured thinking and visual content generation, making complex information easier to understand.
🚀 The Lingguang Flash Application feature allows users to quickly generate personalized tools, such as calculators or planners.
2、xAI launches Grok 4.1, improves quality and speed, free to use!
xAI company launched two models, Grok4.1 and Grok4.1Thinking, significantly reducing the hallucination phenomenon when generating content, and performing well in benchmark tests. However, Grok4.1 may not be the strongest model this year, because Google is preparing to launch Gemini3.0.
【AiBase Summary:】
🌟 Both Grok4.1 and Grok4.1Thinking models are available for free.
📉 The new model reduces hallucinations by three times when generating content, improving accuracy.
🏆 Preliminary tests show that Grok4.1 (Thinking) received the best score, outperforming the previous version.
Details link: https://data.x.ai/2025-11-17-grok-4-1-model-card.pdf
3、Poe AI's group chat function is officially launched: 200 people collaborate, AI model mixing, revolutionizing multi-user interaction experience!
Poe launched a group chat function, supporting 200 people online at the same time with multiple AI models collaboration, suitable for travel planning, creative brainstorming, etc. This function integrates multimodal AI, achieving cross-device synchronization, improving team collaboration efficiency, and promoting the democratization of AI.
【AiBase Summary:】
🧠 Seamless collaboration, any AI dancing together
🌐 Multimodal AI coverage, compatible with over 200 models
📱 Cross-device synchronization, real-time uninterrupted
Details link: https://poe.com/GuessTheFacts
4、Ready to use! OceanBase released its first AI database seekdb, supporting AI-native hybrid search
OceanBase released its first AI database seekdb, which supports AI-native hybrid search, capable of unified hybrid search for vector, full-text, scalar, and spatial geographic data, and is compatible with various mainstream AI frameworks. It aims to reduce the engineering threshold for AI applications and improve development efficiency.
【AiBase Summary:】
✨ OceanBase released its first AI database seekdb, realizing AI-native hybrid search capabilities.
🚀 Supports unified hybrid search for vector, full-text, scalar, and spatial geographic data.
🔧 Ready to use, compatible with more than 30 mainstream AI frameworks, reducing the engineering threshold for AI applications.
5、Domestic AI model Kimi K2 successfully connected to Perplexity, taking an important step forward
The article introduces the domestic AI model Kimi K2 successfully connecting to the globally renowned AI search application Perplexity, becoming the only domestic model to connect. This marks further recognition of domestic AI technology on the international stage, and it was connected at the same time as OpenAI's GPT-5.1, showing the strong competitiveness of domestic models. Perplexity, a rapidly rising AI search application, has a large number of users and a high valuation, while Kimi K2, the latest open-source hybrid expert large model launched by Moonshot, has powerful performance and wide application scenarios.
【AiBase Summary:】
🌟 Kimi K2 Thinking model successfully connected to Perplexity, becoming the only domestic model.
📈 Perplexity attracted 30 million users in just two years, with a valuation exceeding 20 billion dollars.
🚀 Kimi K2 is the latest open-source hybrid expert large model, with great application potential.
6、Google DeepMind releases SIMA 2: A new general agent, competing in complex 3D virtual worlds
Google DeepMind released SIMA2, a general agent based on the Gemini model, aiming to test its performance in complex 3D game worlds. SIMA2 not only understands language instructions but also processes voice, graphics, and emojis, and continuously improves through self-learning. Task completion rate increased to 62%, close to human player level.
【AiBase Summary:】
🌟 SIMA2 integrates the Gemini2.5Flash Lite model, enhancing reasoning and planning capabilities.
📈 Task completion rate increased to 62%, close to human player level, with significant performance improvements.
🛠️ Through self-improvement mechanisms and Genie3 environment generation, it demonstrates adaptability and generality.
Details link: https://deepmind.google/blog/sima-2-an-agent-that-plays-reasons-and-learns-with-you-in-virtual-3d-worlds/
7、ElevenLabs explosive update: image, video, music one-stop generation
ElevenLabs launched a new Image & Video platform, integrating image generation, video generation, voice synthesis, music creation, and sound design, becoming a one-stop AI content factory. The platform not only integrates a global top-tier multimodal model matrix, but also deeply optimizes for creators and marketers, improving the efficiency and quality of content production.
【AiBase Summary:】
🖼️ The new platform integrates image generation, video generation, voice synthesis, music creation, and sound design, achieving one-stop AI content production.
🔊 Integrates a global top-tier multimodal model matrix, including Google Veo, OpenAI Sora, etc., enhancing visual and auditory effects.
💼 Optimized for commercial use, supports multiple aspect ratios, multilingual voiceovers, and audio-video synchronization adjustments, suitable for advertising campaigns.
Details link: https://elevenlabs.io/image-video
8、Kunlun Wanzhi released the lightweight multimodal agent Skywork R1V4-Lite, opening a new era of intelligent interaction
Kunlun Wanzhi's Skywork R1V4-Lite is a lightweight multimodal agent that integrates visual operation, reasoning, and planning capabilities. It can actively perform image operations, external tool calls, and multimodal deep research, significantly improving application flexibility in complex scenarios. Users just need to take a photo, and the system can automatically complete the task without complex prompts.
【AiBase Summary:】
📸 Skywork R1V4-Lite has visual operation capabilities, automatically judging spatial positions, magnifying blurred text, etc.
🔍 Performs well in multimodal understanding benchmark tests, showing strong cross-modal reasoning and knowledge expansion capabilities.
💡 Supports online search and proactive task planning, able to generate executable task chains to provide accurate solutions.
Details link: https://github.com/SkyworkAI/Skywork-R1V
