3/29 AI Express: Manus AI Plans Huge Funding Round / GPT-4o Features Updated / Claude’s ‘Thinking’ Method
AI News Letter
📌 Table of Contents:
💰 Manus AI Plans Huge Funding, Valuation Soars
👁️ Alibaba Launches Qwen’s Latest Visual Reasoning Model, QVQ-Max
🧠 World’s First AI Psychotherapist Report Released
🔄 OpenAI GPT-4o Features Updated
🔬 Anthropic Reveals Claude Model’s “Thinking” Method
🔍 AI Developer Tool: Deep Research Integrated into Code Editors
PART 01
Manus AI Plans Huge Funding Round

Highlights:
- Valuation Soaring:
Manus AI is negotiating funding up to 3.7 billion RMB, with its valuation surging fivefold, demonstrating strong investor confidence. - Commercial Value:
Manus AI has officially announced pricing plans and launched its app on overseas App Stores, accelerating its commercialization.
Pricing Plans:
- Manus Starter: 3,900 credits/month; up to 2 simultaneous tasks; exclusive resources; extended context length; priority access during peak hours.
- Manus Pro: 19,900 credits/month; up to 5 simultaneous tasks; supports high-investment mode and additional beta features; other features same as Starter.
- Overseas Investment Opportunities:
Despite potential international restrictions, U.S. investors continue to show significant interest in Chinese AI products, especially after the rise of companies like DeepSeek.

Industry Insights:
- High-quality technology and products attract global capital despite international competition and restrictions.
- Successful commercialization models are crucial for the long-term development of AI products. Manus AI’s pricing strategy deserves attention.
- American investor interest reflects the rapid development of Chinese AI technology.
Website: Manus AI
PART 02
Alibaba’s Qwen Latest Visual Reasoning Model, QVQ-Max

Highlights:
- Advanced Visual Reasoning:
QVQ-Max not only identifies images but also performs complex visual information analysis and reasoning, including solving mathematical problems and generating code. - Adjustable ‘Thinking’ Mechanism:
The model includes an adjustable ‘thinking’ mechanism to improve accuracy by increasing “thinking time.” - Future Vision:
Plans include developing full visual agents capable of operating devices and playing games, showcasing ambitions in visual AI.
Industry Insights:
- Developers and operators should consider model interpretability and control (like adjustable “thinking” mechanisms) key to improving AI performance and reliability.
- General users can anticipate smarter AI in visual tasks, significantly benefiting fields like autonomous driving and smart homes.
Website: Aliyun Qwen
PART 03
World’s First AI Psychotherapist Report

Highlights:
- Effective Results:
The AI chatbot Therabot significantly alleviated symptoms of depression, anxiety, and eating disorders in clinical trials, performing comparably to traditional cognitive therapy. - Proactive Interaction:
Therabot, trained in cognitive behavioral therapy, proactively initiates conversations and identifies suicide risks.
Industry Insights:
- AI shows significant potential in mental health, effectively complementing traditional therapeutic methods.
- Developing empathetic, knowledgeable AI mental health tools requires extensive psychological research.
- AI could become a vital mental health resource, especially for underserved populations.
Website: Therabot
PART 04
GPT-4o Features Updated

Highlights:
- Enhanced Multi-instruction Parsing: Improved understanding of complex multi-step instructions.
- Technical Problem-solving Boosted: Better handling and solution generation for complex technical and coding issues.
- Improved Logic and Creativity: Advanced innovation and interdisciplinary thinking.
- Streamlined Interaction Interface: Reduced emoji use for optimized professional interactions.
Industry Insights:
- GPT-4o ranked second on the latest LMArena benchmark, surpassing GPT-4.5, with scores 30 points higher than its January performance.
- Updates are immediately available to paid users (e.g., ChatGPT Plus), with general users gaining access in the coming weeks. Developers can access via the chatgpt-4o-latest API, with further rollouts planned.
Website: OpenAI
PART 05
Claude Model’s “Thinking” Method Revealed by Anthropic

Highlights:
- “AI Microscope”: Researchers developed a tool revealing the internal ‘circuits’ of Claude, illustrating its input-output reasoning.
- Universal “Thought Language”: Claude processes concepts similarly across languages like English, French, and Chinese.
- Strategic Word Planning: Claude strategically plans rhymes and word choices in poetry composition.
- Hallucination Prevention Default: A discovered default setting blocks speculation unless overridden by strong confidence, explaining hallucination prevention mechanisms.
Industry Insights:
- Deep understanding of AI models’ internal processes is essential for reliability, safety, and control, promoting trust in AI systems.
- Awareness of AI’s deceptive nature and internal processes becomes critical as more powerful AI systems integrate globally.
Report Link: Anthropic Research
PART 06
Deep Research Integrated into Code Editors

Highlights:
- Developers can directly access real-time web information within their coding environment, boosting efficiency.
- Compatible with popular AI code editors such as Cursor and Windsurf.
Industry Insights:
- Continuous enhancement of AI development tools dramatically improves efficiency and innovation. Integrated, feature-rich development environments are becoming the norm.
Project Link: GitHub Mendable AI
AI Technology News Summary
China’s AI market offers substantial opportunities, particularly in multimodal and healthcare domains. Ongoing evolution in large language models emphasizes the importance of understanding their internal mechanisms. Embracing advanced developer tools to accelerate innovation is key. Staying informed and aligned with cutting-edge developments remains crucial for AI professionals.
ZC Digitals
🚀 Leading digital transformation for enterprises, shaping industry futures. Specializing in AI-integrated customized digital systems, we enable intelligent business process upgrades and deep integrations. Our core team from MIT and Microsoft empowers you with robust AI-driven infrastructure, enhancing efficiency, fostering innovation, and securing industry leadership.