Singapore’s Agnes AI emerges as SE Asia’s sovereign tech contender

By Francis Allan L. Angelo

SINGAPORE — In a region grappling with questions of technological dependence, a Singapore-grown artificial intelligence (AI) startup is redefining what it means to build sovereign digital infrastructure — and doing so at speed.

Agnes AI, founded by Bruce Yang, a Raffles Institution alumnus and National University of Singapore (NUS) PhD, has rapidly scaled into one of Southeast Asia’s most promising AI companies.

Since its official launch on July 4, 2025, the platform has amassed over 5 million users, 200,000 daily active users, and nearly 3 million monthly active users globally. Nearly 50 percent of its user base is concentrated in Southeast Asia.

“Our strategic foundation is Southeast Asia, aiming for over 40% of users in the region, and reaching 300 to 500 million monthly active users globally,” said Yang, AgnesAI’s co-founder and CEO, in an exclusive interview with Daily Guardian.

“We hope Agnes is among the top ten AI companies in terms of users, market impact, brand value, and model capabilities.” 

A homegrown model in a global race

At the core of Agnes AI is SeaLLM-8B, an 8-billion-parameter language model developed entirely in-house.

Unlike many competitors built on U.S.- or China-based open-source foundations, SeaLLM-8B was trained, optimized, and deployed locally.

It has since been open-sourced and is now available on Hugging Face, a gesture that reflects Agnes’s commitment to transparency and ecosystem building.

Despite its relatively smaller size, SeaLLM-8B rivals — and in some tests, outperforms — 20B-parameter models across multiple benchmarks.

Its competitive edge lies in its design, optimized for Southeast Asian languages such as Tagalog, Bahasa Indonesia, and hybrid linguistic patterns like Taglish and Singlish.

“Most large models today support multiple languages, but Southeast Asian languages are still underrepresented in their training data,” said Yang.

“This affects how well they handle code-switching patterns. Agnes-SeaLLM-8B is different because a much larger portion of its training data comes from Southeast Asian languages.”

Deep cultural understanding

Yang emphasized that SeaLLM-8B is not just linguistically adept, it is also culturally grounded.

When asked how Agnes differs from U.S.-centric models like GPT-4, he cited how Agnes responds to questions about traditional Southeast Asian art forms with more narrative depth and cultural sensitivity.

“Agnes does not merely describe shadow puppetry in abstract or historical terms. It understands and reflects how these traditions are embedded in local religious evolution,” Yang said.

“Compared to GPT’s more encyclopedic and analytical style, Agnes uses locally grounded language and context.”

This approach has resonated with Southeast Asian users.

In the Philippines, for example, now one of Agnes’s top markets, educators have adopted the platform’s “AI Slides” tool to create lesson plans, while freelancers and virtual assistants use its productivity features to streamline daily work.

“We were pleasantly surprised by how teachers are actively using AI Slides to create teaching materials,” Yang said.

“It shows strong demand for AI tools that make everyday work more efficient and creative.” 

Built for Southeast Asia — and beyond

Agnes markets itself as an all-in-one AI assistant, integrating research, design, presentations, and communication into a single platform.

Its interface includes modules like Deep Research, AI Slides, AI Sheet, and CoVibe, a group chat feature.

With top-10 productivity rankings on Google Play in Singapore, the Philippines, Indonesia, and Vietnam, the company is now preparing for global expansion.

“We aim to create a major AI platform — something like Grab or Shopee but in the AI space,” Yang said.

“The Southeast Asian AI market isn’t as crowded as the U.S. or China, and each niche has a clear leader.”

To support this growth, Agnes is preparing to train a next-generation model locally in collaboration with NUS and NTU, while nearing the close of a funding round reportedly worth tens of millions of U.S. dollars.

Yang confirmed the funding would support both scaling and international expansion. 

Open source, yet sovereign

In contrast to the growing trend of closed AI ecosystems, Agnes has opted for a hybrid approach — releasing its core 8B model as open-source, while keeping its larger models and commercial platforms proprietary.

“Our approach to sovereign AI is built around a complete ecosystem — from the model to the framework, top-level applications, and user data,” Yang said.

“Open-sourcing is a way to develop this ecosystem, allowing more people to participate, use, and innovate.”

He noted that open-sourcing does not erode their commercial moat.

“Our core value lies in user traffic and brand influence. Even with open-source models, deployment and building end-user applications are still required, so our key technology remains under our control.”

Bridging the talent gap

Agnes is also contributing to Southeast Asia’s AI ecosystem by reversing the brain drain. Yang himself returned from Silicon Valley during the COVID-19 pandemic to pursue his PhD and build Agnes in Singapore, assembling a team with talent from local universities and global institutions, including MIT, Stanford, and UC Berkeley.

“Southeast Asia still has some gaps compared to the U.S., but that can be an advantage,” he said.

“There are relatively few AI companies in the region, so as a leading regional AI company, Agnes can attract more local and global talent.”

He acknowledged, however, that the talent gap is still most severe in product-level model engineering, infrastructure, and fundamental AI research — areas where the region must catch up to remain competitive.

Competing with tech giants — and staying independent

With rivals like Microsoft Copilot and Google Gemini backed by near-limitless capital, Agnes faces stiff competition.

But Yang remains focused on building a differentiated product that meets regional needs.

“There are many strong AI products on the market, each with a different positioning. Our focus is to do well in specific areas where we can make the most impact within our market,” he said.

The company aims to go public by 2027 and is currently not pursuing acquisitions, although Yang did not rule it out in the long term.

For now, the emphasis is on scaling users and model capability while retaining independence.

High ambitions, higher stakes

Ultimately, Yang views Agnes’s journey as more than a business — it’s a regional bet on cultural, economic, and technological self-determination.

“AI sovereignty starts with cultural confidence,” he said.

“Beyond hardware like chips, the entire AI framework needs to be supported by local systems. Agnes is built with Southeast Asia as the priority.”

And if it fails?

“We hope Singapore and the region will learn that world-class AI can be built here — but only if there’s long-term commitment,” Yang said.

“Our goal is to be the leading AI company in the region, and to list on Nasdaq as a Singapore-based company.”

Whether that vision becomes reality, Agnes already stands as a proof point that Southeast Asia can shape, not just consume, the future of AI.