Manus AI – Will China Continue to shake the AI world?

Following the debut of the DeepSeek R1 model at the end of January, a Chinese startup has once again stirred the tech world with Manus AI – the world’s first AI Agent that is expected to make breakthroughs similar to DeepSeek.

Manus is the first fully autonomous AI Agent in the world, capable of independently handling complex tasks such as: planning travel itineraries, conducting in-depth stock analysis, researching real estate, filtering job applications, and more.

Manus AI impresses with several standout features:

+ Comprehensive automation: Manus functions like a manager, overseeing multiple smaller agents. Upon receiving a task, Manus analyzes the request, creates a plan, monitors the agents’ progress, and ultimately delivers a final answer to the user.

+ Cloud-based operation: Manus can continue working even when the user turns off their computer, and only notifies the user when the task is completed.

+ Integration of powerful models: Manus combines models such as Claude 3.5 Sonnet by Anthropic, Qwen by Alibaba, and 29 other open-source AI tools to efficiently perform complex tasks.

The emergence of Manus further highlights China’s rapid AI development, demonstrating not only an ambition to lead in the AI field but also opening up new opportunities to apply AI in everyday life and work.

2. DeepSeek surprises the world with new AI model that outpaces competitors

On March 25, DeepSeek released an upgraded version of its DeepSeek V3 model, marking a major step forward in the global AI race.

The new model, DeepSeek-V3-0324, stands out thanks to its mixture-of-experts (MoE) architecture, which allows it to activate only about 37 billion parameters out of a total of 685 billion for each task. This significantly conserves computing resources while maintaining performance on par with larger models. Additionally, DeepSeek integrates two advanced technologies: Multi-Head Latent Attention (MLA) and Multi-Token Prediction (MTP), enabling the AI to better understand content and deliver answers up to 80% faster.

Notably, DeepSeek V3 can run directly on high-end consumer devices like the Mac Studio M3 Ultra without the need for a supercomputer, making cutting-edge AI more accessible to a wider audience.

By releasing an advanced model like DeepSeek-V3-0324 for free, DeepSeek not only increases pressure on OpenAI but also signals a future where the AI gap between China and the U.S. is rapidly narrowing.

3. ChatGPT “unlocks” new image generation feature

OpenAI has officially integrated a new image generation feature into ChatGPT, named Images in ChatGPT.

This new feature is now available to both free users and subscribers of the Plus, Pro, and Team plans, allowing users not only to chat but also to generate highly realistic images using simple text prompts.

A notable improvement in this feature is the use of "Binding" technology, which ensures precise alignment between object attributes and the objects themselves in the image—ensuring the AI accurately understands and visualizes user requests. Additionally, the ability to generate readable text within images has been significantly upgraded. ChatGPT can now produce clear, coherent text in images, eliminating the distortions commonly seen in many AI-generated visuals.

Although image generation within ChatGPT may be slower than in other models, OpenAI representatives said it is a “worthwhile trade-off” in exchange for higher image quality, and that their team is working to improve this limitation.

4. Google creates buzz with hyper-realistic image AI and the “Smartest model in the world”

In mid-March, Google unveiled Gemini 2.0 Flash Experimental - an AI tool that allows users to create and edit images using natural language commands. The tool immediately caught the tech world's attention for its ability to perform a wide range of tasks, such as modifying image details, removing backgrounds, merging photos, adding text, and restoring old images.

Despite some limitations, such as lower resolution and occasional distortion or inaccuracy in text rendering - Gemini 2.0 Flash Experimental has been praised as an impressive image generation tool, with the potential to compete with, or even replace, professional editing software like Adobe Photoshop in the near future.

By the end of March, Google made another splash by launching Gemini 2.5 Pro, a new AI model hailed as “the smartest in the world”. The model reportedly surpasses many current AI systems in areas like reasoning, mathematics, and science. Notably, Gemini 2.5 Pro is capable of inference before responding, boosting its performance and accuracy when handling complex tasks.

In a surprising move, just days after the launch, Google announced that Gemini 2.5 Pro would be made available for free—demonstrating the tech giant’s commitment to democratizing AI and accelerating its integration into daily life.

5. NVIDIA asserts global AI leadership at GTC 2025

The NVIDIA GTC 2025 Conference - one of the most anticipated and prominent tech events of the year officially took place from March 17 to 19.

At the event, NVIDIA CEO Jensen Huang unveiled a series of groundbreaking innovations in the fields of AI and high-performance computing. Among them, the Blackwell Ultra GPU, which boasts 1.5 times the AI performance of the current Blackwell GPU, is expected to launch in the second half of this year.

Following the Blackwell Ultra, NVIDIA plans to release the Rubin GPU in the second half of 2026, promising even better AI training and inference performance compared to the Grace-Blackwell chipset. In addition, the company introduced GR00T N1 - its first open-source foundation model for humanoid robots, which is expected to make its debut at NVIDIA’s theme parks next year.

GTC 2025 showcased NVIDIA’s long-term strategy to redefine its role in the AI industry: not only as a hardware provider but also as a builder of a comprehensive AI infrastructure ecosystem, accelerating the development and application of AI across diverse sectors.

6. MWC 2025 - The world’s premier mobile conference officially kicks off

Mobile World Congress 2025 (MWC 2025), the world’s largest technology and telecommunications event, has officially opened, bringing together leading global tech corporations such as Samsung, Huawei, Xiaomi, and Nokia.

Under the theme "Unite. Connect. Create.", MWC 2025 emphasizes the role of technology in promoting sustainable development and global innovation. The event is not only a stage for showcasing cutting-edge tech products but also a platform for business networking, expanding partnerships, and sharing the latest technology trends.

Notably, Viettel Group—the only representative from Vietnam presented 22 technology products spanning 5G infrastructure, digital platforms, and AI applications, affirming Vietnam’s technological capabilities on the international stage.