Welcome to the "AI Daily" column! 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://top.aibase.com/
1. Tencent Hunyuan opensources and releases 0.5B, 1.8B, 4B, 7B models
Tencent Hunyuan team launched four small-scale open-source models suitable for consumer-grade GPUs and low-power scenarios, supporting cost-effective fine-tuning in vertical fields. These models excel in reasoning speed, cost-effectiveness, and long-text processing capabilities and have been launched on multiple open-source platforms.
【AiBase Summary:】
✨ Four small-scale models are designed for consumer devices, suitable for various low-power scenarios.
🚀 The models have fast reasoning and long-text processing capabilities, capable of handling extremely long content at once.
🔧 Support multiple deployment methods, suitable for diverse needs from edge to cloud.
More details: https://hunyuan.tencent.com/modelSquare/home/list
2. Kunlun Wanzhi releases and opensources the new inference large model MindLink
Kunlun Wanzhi released and open-sourced the latest inference large model Skywork MindLink. The model achieves dynamic path selection through an innovative inference framework, improving answer transparency and efficiency, and achieving excellent results in multiple evaluations.
【AiBase Summary:】
🧠 Skywork MindLink adopts a plan-based reasoning new paradigm, optimizing multi-turn dialogue experience.
🏆 Performs excellently in multiple benchmark tests, winning multiple gold medals in math competitions.
🔧 Built-in adaptive reasoning system, automatically adjusts generation strategy according to task difficulty.
More details: https://github.com/SkyworkAI/MindLink
3. Bilibili launches AI voice translation function: retaining the original voice tone of UPs, solving the problem of anime culture going global
Bilibili launched its self-developed AI voice translation function to solve the content interoperability issue after merging the international and domestic versions. The technology can retain the original voice tone, voice characteristics, and speaking habits of UPs, providing overseas users with a more natural Chinese content experience.
【AiBase Summary:】
✅ Bilibili launched AI voice translation function, supports English translation, retains the original voice tone and voice of UPs.
🔄 Uses adversarial reinforcement learning and Deep Research technology to ensure accurate translation while preserving cultural nuances.
🌐 Will expand support for more languages such as Japanese in the future, helping Bilibili's internationalization strategy.
4. Google Gemini 2.5 Deep Think released! IMO gold medal added, can AI new king reshape the future?
The Gemini 2.5 Deep Think model released by Google DeepMind shows outstanding reasoning ability in multiple fields, especially winning a gold medal in the International Mathematical Olympiad in 2025. The model introduces parallel thinking and reinforcement learning technologies, enhancing complex task processing capabilities, and performs well in coding and cross-domain knowledge tests.
【AiBase Summary:】
🧠 Introduces parallel thinking mechanism, enhances complex problem-solving capabilities.
🏆 Won gold medal in IMO competition, demonstrating top-level mathematical reasoning skills.
🚀 Supports multimodal and long context, applicable to various scenarios.
5. OpenAI CEO showcases GPT-5 new features, can efficiently integrate online information
OpenAI CEO Sam Altman shared chat records of GPT-5 on social media, demonstrating its powerful information integration capability. GPT-5 gave a positive evaluation of the sci-fi animation "Pantheon" and mentioned that the show received 100% positive reviews on Rotten Tomatoes. This event marks the first public appearance of GPT-5, causing widespread attention in the tech industry.
【AiBase Summary:】
🌟 GPT-5 made its first public appearance, demonstrating its powerful information integration capability.
📺 OpenAI CEO recommended the sci-fi animation "Pantheon" and shared the model's positive evaluation.
🔍 GPT-5's rating on the Rotten Tomatoes website is "100% critic approval", causing widespread attention.
6. Apple forms AI answer engine team: challenge ChatGPT, may reshape Siri and Safari search experience
Apple has formed a dedicated team to develop AI applications similar to ChatGPT, aiming to improve the search and interaction experience of its core products. The team, named Answers, Knowledge, and Information, focuses on building an answer engine that can use online information to answer user questions.
【AiBase Summary:】
🍎 Apple formed a new team to develop AI applications similar to ChatGPT to enhance search and interaction experiences.
🔍 The answer engine may be a standalone application or integrated into products like Siri and Safari, providing smarter search functions.
🌐 Apple hopes to reduce reliance on third-party AI services and respond to the impact of Google's antitrust case.
7. Gaode Map announces full AI transformation, launching the world's first AI-native map application "Gaode Map 2025"
Gaode Map officially launched the world's first AI-native map application - Gaode Map 2025, marking a major breakthrough in its technological field. The application combines spatial intelligence technology, enhances the intelligence of maps through multimodal information perception, and will have a positive impact in multiple areas.
【AiBase Summary:】
🚀 Gaode Map launched the world's first AI-native map application, achieving a technological breakthrough.
🧠 Spatial intelligence technology enhances the multimodal information perception capability of maps.
🚗 The application will expand to smart cars, smart glasses, and other fields, improving travel efficiency.
8. Adobe Photoshop launches "Harmonize": using AI to automatically match lighting, achieving seamless image compositing
Adobe simplifies complex image editing processes through a series of generative AI tools, such as "Harmonize," improves the efficiency of image compositing and modification, and introduces content credentials to ensure the authenticity of images.
【AiBase Summary:】
🖼️ "Harmonize" tool automatically matches lighting, color, and shadows of image elements to achieve seamless compositing.
🔍 AI-driven image enhancement feature can increase resolution up to 8 million pixels without losing quality.
🔒 Content credentials feature provides a reliable tracking chain for the editing history of images, ensuring the authenticity of digital content.
9. NVIDIA launches Cosmos DiffusionRenderer: revolutionary video rendering technology
NVIDIA launched Cosmos DiffusionRenderer, a new video diffusion framework for high-quality image and video re-lighting and de-lighting. This technology is a significant upgrade to the original DiffusionRenderer, improving rendering quality through improved data planning processes.
【AiBase Summary:】
🌟 The technology is a major upgrade to NVIDIA's original DiffusionRenderer, offering higher quality image and video rendering.
💻 Users need to install Python3.10 and at least 16GB of NVIDIA GPU memory, and create the relevant conda environment.
🎥 Supports de-lighting and re-lighting of images and videos and can render using various environmental light maps.
More details: https://github.com/nv-tlabs/cosmos1-diffusion-renderer
10. Android development revolution! Google Android Studio free Agent mode launched, surpassing Apple ecosystem?
Google announced the launch of the free Agent mode of Android Studio at Google I/O 2025. This mode is based on Gemini 2.5 Pro, improves development efficiency through natural language interaction, supports cross-file tasks, UI code modifications, and custom rules. Its features not only challenge Apple's Xcode ecosystem but also provide developers with more efficient tools.
【AiBase Summary:】
🤖 Agent mode: Based on Gemini 2.5 Pro, completes complex development tasks through natural language interaction.
🔍 Core functions: Supports quick modification of UI code, custom rule settings, and a million-token context window.
🚀 Competitive advantage: Free open Agent mode, directly challenging Apple's Xcode ecosystem.
11. Google opensources structured information extraction tool langextract, which can provide precise source location
Google open-sourced the LangExtract tool, which can efficiently extract structured information from unstructured text, applicable to multiple fields such as medicine, literature, and business, providing developers with powerful solutions.
【AiBase Summary:】
🧠 Accurate source tracing: Extracted results can be mapped to specific positions in the source text, facilitating verification and data accuracy tracing.
🧩 Reliable structured output: Define the output format with a few examples, ensuring compliance with the user's predefined JSON schema.
📊 Interactive visualization: One-click generates HTML reports, intuitively view extraction results, and improve review efficiency.
More details: https://github.com/google/langextract
12. Figma developer mode major update: Colorful annotations and MCP protocol dual upgrades, design to code efficiency soars
Figma recently conducted a comprehensive upgrade of the developer mode, introducing a colorful interactive annotation system and a major improvement in the Model Context Protocol (MCP). These updates significantly enhanced the efficiency of design and development collaboration, setting a new benchmark for the industry.
【AiBase Summary:】
🎨 Colorful interactive annotation system allows designers to mark information with different colors, improving development understanding efficiency.
🔄 MCP protocol upgrade supports transmission of structured data, making AI-generated code more suitable for actual needs.
🚀 New features such as Ready for Dev view simplify the design handoff process, improving team collaboration efficiency.