AI Daily: Claude Opus 4.7 Released; Alibaba Open Sources Qwen3.6-35B-A3B; Perplexity Launches AI Assistant for Mac

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 learn about innovative AI product applications.

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

1、Claude Opus 4.7 Officially Released: Reliability Is More Important Than Intelligence

The release of Claude Opus 4.7 marks a major breakthrough in AI model reliability by Anthropic. It performs well in multiple benchmark tests and has made significant improvements in task execution stability and accuracy.

【AiBase Summary:】

🧠 Claude Opus 4.7 performs excellently in programming benchmark tests, surpassing GPT-5.4 and Gemini 3.1 Pro.

💡 The new 2576-pixel long-side recognition capability improves visual reasoning performance.

🔒 Significant improvement in task resilience, enabling it to bypass tool chain crashes autonomously.

2、Alibaba Open Sources Qwen3.6-35B-A3B: A 30 Billion Parameter Model Achieves a Leap in Programming Capabilities

On April 16, 2026, Alibaba's Qwen large model team open-sourced the sparse mixture-of-experts (MoE) model Qwen3.6-35B-A3B, marking a key breakthrough in lightweight models for intelligent agent programming. The total parameter count of the model is 35 billion, but the activated parameters at runtime are only 3 billion, outperforming dense models with 27 billion parameters and demonstrating capabilities comparable to large-scale models. As a full-modal open-source model, it excels in spatial intelligence and visual perception, and has been integrated into Qwen Studio and provides API services through the Alibaba Cloud BaiLian platform, supporting the chain-of-thought retention function, and is compatible with mainstream AI coding assistants.

【AiBase Summary:】

🧠 The Qwen3.6-35B-A3B model achieves a leap in programming capabilities with 3 billion activated parameters.

🚀 The model surpasses dense models with 27 billion parameters in several core programming benchmark tests.

🌐 The model has been integrated into Qwen Studio and provides API services through the Alibaba Cloud BaiLian platform.

3、OpenAI Updates Codex: Supports Mac Desktop Control, Multi-Agent Parallelism, and Long-Term Task Execution

OpenAI made significant updates to Codex, enhancing its capabilities in Mac desktop application control, multi-agent parallel operation, and long-term task execution. It also optimized browser experience and plugin ecosystem, further improving development efficiency and task processing flexibility.

【AiBase Summary:】

💻 Codex now has the ability to control Mac desktop applications, autonomously operate interface elements, and perform tasks.

🔄 Supports multi-agent parallel operation, improving development efficiency and task handling capabilities.

📅 Introduces memory functionality, enabling long-term task execution and seamless workflow recovery.

4、Opera Launches Browser Connector Feature, Enabling Seamless Integration of AI Chatbots

Opera launched a new feature called "Browser Connector," allowing users to integrate more AI chatbots into their browser, such as OpenAI's ChatGPT and Anthropic's Claude. This feature is based on the MCP protocol, aiming to establish secure bidirectional connections between AI models, external data sources, and search engines. Users can install and connect these AI tools through simple setup steps, thereby enhancing browsing experience and flexibility.

【AiBase Summary:】

🤖 Opera introduces the Browser Connector, allowing users to integrate more AI chat tools into the browser.

🔗 Users can simply install and connect in the settings to use them, enhancing browsing experience and flexibility.

🌟 This feature is applicable to Opera One and Opera GX, supporting AI tools like ChatGPT and Claude.

5、NVIDIA Releases Lyra2.0: Generate 90-Meter 3D Environments from a Single Photo, Surpassing Competitors in Multiple Metrics

NVIDIA's Lyra2.0 technology builds large-scale, highly coherent virtual environments from a single photo, solving the image distortion problem under long-distance camera paths, marking a major breakthrough in AI's 3D space understanding and real-time environment simulation.

【AiBase Summary:】

🧠 Generates 3D environments up to 90 meters from a single photo, enhancing AI's understanding of space.

🔄 Introduces dual innovation strategies to solve traditional video model issues of spatial distortion and error accumulation.

🚀 Achieves seamless integration with physics engines, providing efficient environment support for robot simulation training.

6、Google Gemini Integrates Personal Photo Albums, AI-Generated Images Move Toward True Personalization

Google Gemini integrates personal photo albums, moving AI-generated images toward true personalization.

【AiBase Summary:】

📱 Google Gemini integrates personal photo albums, achieving a personalized experience in AI image generation.

🎨 Using the Nano Banana tool, Gemini can directly access private photos in Google Photos without manual upload.

💡 This update breaks down the barriers between private media and AI generation capabilities, promoting image generation towards truly personalized development.

7、Perplexity Launches Mac AI Assistant: Runs 24/7, Can Operate Computer on Behalf of Users

Perplexity launched the Mac AI assistant Personal Computer, which not only understands commands but can also directly access the file system and native applications, marking a deep advancement of AI interaction from "dialogue box" to "operating system."

【AiBase Summary:】

🧠 Perplexity, a leader in AI search, released the Mac AI assistant Personal Computer, capable of directly accessing the file system and native applications.

💻 The tool supports text and voice interaction and can perceive the user's active window in real time, proactively offering operational suggestions.

🔒 Perplexity designed a strict defense mechanism for the assistant, completing all operations within an isolated security sandbox to ensure user privacy.

8、Take a Photo and It's Automatically Entered, Tags Are Named Automatically! iOS 27's Four AI Upgrades Exposed, Reimagining iPhone Interaction

Apple revealed several new Apple Intelligence features in the iOS 27 Beta version, including nutrition label parsing, paper information recognition, physical card coupon digital scanning, and intelligent tag group naming. These features demonstrate Apple's further deepening of AI technology and lay the foundation for future wearable device ecosystems.

【AiBase Summary:】

🍎 Nutrition Label Parsing: The system can automatically scan food packaging data and synchronize it to health apps.

📄 Paper Information Recognition: Supports extracting phone numbers or addresses from printed materials and adding them to the contact list with one click.

📱 Digital Scanning of Physical Coupons: Converts event tickets and membership cards into digital passes through AI image recognition.