Agnes AI hits 5m users & open-sources SEA-focused LLM

Singapore-based Agnes AI says its proprietary platform has passed 5 million registered users within six months and has entered the Top 10 productivity apps in several Southeast Asian markets.

The company said the service has reached 200,000 daily active users and 2.97 million monthly active users. It said close to half of its users come from Southeast Asia. It also cited Google Play rankings in Singapore, the Philippines, Vietnam, and Indonesia.

Agnes AI launched in July 2025, according to the company. It positions the product as an all-in-one AI assistant.

Product focus

Agnes AI said its platform integrates search, research functions, slide creation, design tools, spreadsheet features, filters, group chat, and a news section. It said the product flows from question and research through to presentation. It also said users can work without switching apps.

The company sits in a regional market where consumer and workplace AI use has expanded quickly, alongside wider debates about where core AI models get trained and who controls them. Governments and companies have put increased emphasis on domestic AI supply chains and local development talent.

Agnes AI said it developed its own large language model in-house in Singapore. The firm said the model has 8 billion parameters and that the team built it from the ground up.

Model release

The company has also open-sourced the model, which it calls Agnes-SeaLLM-8B. It said the release is available on Hugging Face.

Agnes AI said its model supports Southeast Asian languages, alongside Chinese and English. The company described the work as designed for cross-cultural use.

It also claimed benchmark results. The company said the 8-billion-parameter model delivers state-of-the-art performance among sub-10B models. It added that it outperforms several 20B models across benchmarks. It attributed the results to optimisation and training techniques.

The company referenced a paper titled "Stable and Efficient Policy Optimization for Agentic Search and Reasoning (DSPO)". It said the paper was submitted to ICLR 2025 and published on arXiv.

Founder background

The company identified its founder as Bruce and described him as a Raffles Institution alumnus and an AI PhD at the National University of Singapore. It said he studied Mathematics and Computer Science at the University of California, Berkeley.

Agnes AI said Bruce worked at Microsoft and LinkedIn. It also said he co-founded a startup that recorded millions of downloads.

During the COVID-19 pandemic, Bruce returned to Singapore, according to the company. It said he pursued a PhD in AI at the National University of Singapore and began laying the foundations for the Agnes AI project.

Agnes AI said it assembled a team with talent drawn from the National University of Singapore and Nanyang Technological University. It also cited international institutions including MIT, Stanford, and UC Berkeley.

Regional angle

The company framed its approach as a bid for locally controllable AI. It said it trained and developed the system in-house, rather than relying on overseas open-source foundations.

It said this approach links to Southeast Asia's linguistic and cultural context. The company also pointed to what it described as a widening talent gap in product-level model development in Singapore.

Agnes AI also positioned its progress against global moves around domestic AI development. It cited Europe's backing of Mistral and national initiatives in Japan and the Middle East.

Next steps

Agnes AI said it is preparing to train a next-generation, larger-scale model locally. It said it is collaborating with the National University of Singapore and Nanyang Technological University on the effort.

The company also said it is nearing the close of a funding round in the tens of millions of dollars. It said the funding would go towards model scaling and international expansion.