Agnes AI, a next-generation artificial intelligence (AI) agent developed and trained in Singapore, announced Wednesday that it has surpassed 2 million registered global users with approximately 150,000 daily active users since its official launch in July 2025.
The firm said in a statement that about 50 percent of users are from Southeast Asia.
Looking ahead, the firm will continue to translate research into product at pace, pairing an in-house engineering team with academic collaborators so ideas move rapidly from lab to everyday use.
The company is also preparing to train a larger next-generation model in Singapore with the National University of Singapore and Nanyang Technological University, while engaging universities, government agencies, and corporates across Southeast Asia, and exploring integrations with regional technology providers to embed its agentic workspace in both education and workplace ecosystems.
Founded by Bruce Yang, Agnes AI is an all-in-one AI assistant, integrating search, deep and wide research, real-time slide generation, design tools, and shared workspaces so users can move from research to content to presentation without switching apps.
It runs on a fully proprietary, locally controlled technology stack that the company designs and maintains end to end, marking a concrete step toward Singapore’s goal of locally controlled, sovereign AI infrastructure.
The firm is led by Founder Bruce Yang alongside Evan Pu, Professor of Computer Science at Nanyang Technological University, Xiaofan Li, Assistant Professor at the National University of Singapore School of Computing and Linus Lee, a Singaporean AI researcher and Stanford graduate.
Instead of stitching together overseas open-source models, Agnes AI built its own architecture from the ground up.
The result is Agnes-R1, a seven-billion-parameter model designed for search, research and presentation workflows.
In testing on real commercial tasks, the model has shown faster reasoning, lower compute costs per task and higher output quality.
On standard question-answering benchmarks, the Agnes-R1 7B model outperformed similar systems by 34.1 percent on average and exceeded some earlier 14B models by nearly 9 percent on complex multi-step reasoning, while maintaining stable training and consistent results.
“Singapore is investing to build AI that we can understand, control, and trust. Agnes AI fits that architecture by combining a proprietary model, locally operated infrastructure, and a product that ships measurable outcomes in classrooms and boardrooms,
“When you can host, inspect, and improve your own AI, you are not just consuming technology. You are strengthening the country’s sovereign AI backbone,” said Yang.
Singapore’s Bluwhale secures $10M strategic Series A in institutional funding