AI Singapore

AI Singapore (AISG) is a national artificial intelligence research and development programme established in 2017 by Singapore's National Research Foundation and hosted at the National University of Singapore. The programme operates as Singapore's principal sovereign-AI vehicle and has produced the SEA-LION family of open-weights large language models specialized in Southeast Asian languages, including English, Mandarin Chinese, Indonesian, Vietnamese, Malay, Thai, Burmese, Lao, Filipino, Tamil, and Khmer. AI Singapore's mission combines national talent development through programmes including the AI Apprenticeship Programme, industry-AI deployment through the 100 Experiments programme, and Southeast-Asian-language model development through SEA-LION.

At a glance

• Founded: May 2017 in Singapore by the National Research Foundation as a five-year, S$150 million national AI programme. Renewed and expanded multiple times through 2026.
• Status: Government-funded national programme. Hosted at the National University of Singapore. Operates as a research-and-deployment organization rather than a commercial company.
• Funding: Singapore government funding through the National Research Foundation. Original 2017 commitment of S$150 million; subsequent expansions including the National Multimodal LLM Programme contributing additional government funding.
• CEO: Programme leadership operates through a Deputy Executive Chairman structure. Professor Mohan Kankanhalli (Deputy Executive Chairman, Talent and Ecosystem) and Laurence Liew (Director, AI Innovation) lead operational direction.
• Other notable leadership: Leslie Teo (Senior Director of AI Products; SEA-LION programme leader). Programme leadership coordinates with the Singapore government's Smart Nation initiative and the Infocomm Media Development Authority (IMDA).
• Open weights: Yes. The SEA-LION family is released open-weights through Hugging Face under permissive licensing. Selected research artifacts (datasets, evaluation suites) are also publicly available.
• Flagship models: SEA-LION v4 (2026 family covering 7-billion through 70-billion parameter variants), SEA-LION-Embedding suite (March 2026), regionally-aware Vision-Language models in development.

Origins

AI Singapore was launched in May 2017 by the National Research Foundation as part of the Singapore government's broader Smart Nation initiative. The programme was framed at launch as a five-year, S$150 million national commitment to artificial intelligence research, talent development, and industry-AI deployment. Singapore's strategic position as a small, technology-focused nation with deep international research collaboration made AI Singapore an unusual organizational structure: a publicly funded national programme rather than a commercial company or a single research institute.

The programme's original focus areas combined research grants for university AI research, the 100 Experiments programme to match AI researchers with Singaporean industry partners, and the AI Apprenticeship Programme to train AI engineers. Through 2018 to 2022, the programme contributed to Singapore's AI talent base and produced research outputs across multiple Singapore-based universities.

The SEA-LION programme emerged in 2023 as the response to the global large-language-model commercial wave. Singapore's strategic position as the principal Southeast-Asian regional hub combined with the regulatory and cultural importance of capable AI systems for languages underrepresented in US-frontier-lab training data motivated the open-source SEA-LION line. The first SEA-LION releases in 2024 covered text models in 7-billion, 8-billion, 9-billion, and 70-billion parameter scales, with explicit fine-tuning for Southeast Asian languages.

In 2024 and 2025, AI Singapore expanded SEA-LION through collaborations with Google DeepMind (the Gemma 2-based SEA-LION variants), with Meta AI / FAIR (Llama-based variants), and with regional research partners across Southeast Asia. The Southeast Asian Languages in One Network Data (SEALD) initiative provided training-data curation across the 11 supported languages.

The 2026 release sequence has expanded the SEA-LION family. The SEA-LION-Embedding suite released in March 2026 provided regional retrieval-and-search capability. The SEA-LION v4 family announced in 2026 includes small efficient open models alongside regionally-aware Vision-Language models in development. A January 2026 collaboration with Dell Technologies announced optimization of SEA-LION for Dell AI PCs and edge infrastructure.

Mission and strategy

AI Singapore's stated mission is to develop Singapore's AI ecosystem through research excellence, talent development, industry deployment, and regionally-aware model development. The strategy combines a national-AI-program logic (talent, ecosystem, research) with a sovereign-AI-product logic (SEA-LION as the open Southeast-Asian-language AI capability layer).

The strategy combines four threads. First, fundamental research through grant funding to Singapore-based university AI research, contributing to global AI research output and Singapore-based research capacity. Second, talent development through the AI Apprenticeship Programme (AIAP), which trains AI engineers through structured industry placements, and through educational outreach. Third, industry deployment through the 100 Experiments (100E) programme, which matches AI researchers with industry partners on real-world AI projects. Fourth, the SEA-LION model line as the principal sovereign-AI product output, providing capable open-weights AI for Southeast Asian languages and serving as a foundation for regional industry-and-government deployment.

The competitive premise of SEA-LION is that US frontier labs and Chinese AI labs underweight Southeast Asian languages in their training data and product priorities, creating a structural opening for a regionally focused open-weights model. The programme's national-funding structure allows long-horizon investment without commercial-revenue pressure, an unusual position among AI organizations of comparable capability.

Models and products

• SEA-LION v4 family. 2026 release sequence covering small efficient open models alongside larger flagship variants. Targeted at Southeast-Asian-language tasks across the 11 supported languages.
• SEA-LION-Embedding suite. Released March 2026. Regional retrieval-and-search embedding models tuned for Southeast Asian languages.
• SEA-LION (7B, 8B, 9B, 70B). 2024 to 2025 release sequence. Open-weights text models with Southeast-Asian-language fine-tuning.
• SEA-LION based on Gemma 2. Collaboration variants built on Google DeepMind's Gemma 2 base model.
• SEA-LION based on Llama. Collaboration variants built on Meta AI / FAIR's Llama base.
• SEALD (Southeast Asian Languages in One Network Data). Training-data curation programme.
• SEA-LION Vision-Language models. In development as of 2026, expanding the family into multimodal capability.

The principal distribution channel is Hugging Face for open-weights releases. AI Singapore also collaborates with industry partners (Dell, Google DeepMind, Meta AI / FAIR, regional research organizations) for downstream deployment and capability extensions.

Benchmarks and standing

SEA-LION's benchmark profile emphasizes Southeast-Asian-language evaluation rather than the standardized English-language leaderboards (Artificial Analysis Intelligence Index, LMArena, GPQA Diamond). Industry coverage and academic evaluation have characterized the SEA-LION family as the leading open-weights option for Southeast-Asian-language tasks at comparable model scales.

The programme's standing in the global AI landscape rests on the SEA-LION open-weights distribution, the Singapore government's strategic-AI commitment, the talent-development outputs through AIAP, and the academic-research collaborations across Singapore and Southeast Asia.

The SEA-LION family is not represented at the top of global frontier-tier benchmarks given its specialized regional focus and smaller-scale model investment relative to commercial frontier labs. The programme's competitive frame is regional rather than global frontier capability, and its commercial impact is measured through SEA-LION adoption by Southeast Asian developers and enterprise customers.

Leadership

As of April 2026, AI Singapore's senior programme leadership includes:

• Professor Mohan Kankanhalli, Deputy Executive Chairman (Talent and Ecosystem). Provost's Chair Professor of Computer Science at the National University of Singapore and Director of the NUS AI Institute. Senior academic-research leadership for the programme.
• Laurence Liew, Director, AI Innovation. Singapore-AI-and-technology leader. Leads the 100 Experiments programme and the AI Apprenticeship Programme. Public face for AI Singapore on industry-AI engagement.
• Leslie Teo, Senior Director of AI Products. Programme leader for SEA-LION and the related Southeast-Asian-language model development.

The programme operates through a distributed structure with research grants and collaborations across Singapore-based universities (including the National University of Singapore, Nanyang Technological University, and Singapore Management University) and with regional and international research partners.

Funding and backers

AI Singapore's funding structure is Singapore government funding through the National Research Foundation. The original 2017 commitment was S$150 million over five years, with subsequent renewals and expansions including the National Multimodal LLM Programme. Specific 2026 funding levels have not been broadly publicized but are reported as expanded relative to the original 2017 commitment.

The programme's government-funded structure distinguishes it from commercial AI organizations and supports long-horizon investment in talent development, regional language AI capability, and research grants without near-term commercial revenue pressure. The Singapore government's broader Smart Nation initiative and the IMDA's emerging-technologies programmes provide additional adjacent funding for AI deployment in Singapore.

Industry position

AI Singapore occupies a structurally distinctive position in the global AI landscape. The combination of national-AI-program funding stability, the Singapore-based research and academic collaborations, and the SEA-LION open-weights regional-language model line produces a profile that no other sovereign-AI organization in Southeast Asia matches and few comparable national programmes globally match across all dimensions.

Industry coverage has frequently characterized AI Singapore as a model for national-AI programmes in non-frontier-lab nations, with the SEA-LION line specifically cited as evidence that capable regional-language AI can be produced through public funding and academic-industry collaboration without commercial frontier-lab structure.

Strategic risks include the dependence on continued Singapore government funding commitment, the operational complexity of programme-organization structure relative to a single commercial entity, and the open question of whether SEA-LION can keep pace with frontier-lab capability advances as base-model frontiers continue to scale. Strategic strengths include the funding stability, the academic-research integration, the Southeast-Asian-language specialization, and the international collaborations with Google DeepMind, Meta AI / FAIR, and regional partners.

Competitive landscape

AI Singapore competes with and collaborates with several AI labs:

• Google DeepMind and Meta AI / FAIR. Frontier-lab collaborators on Gemma- and Llama-based SEA-LION variants. Also competitors at the global frontier-model tier.
• Mistral AI, Aleph Alpha, and other European sovereign-AI providers. Peer national-and-regional sovereign-AI organizations.
• Krutrim and Sarvam AI. Indian sovereign-AI peers focused on Indic languages.
• Naver Cloud and other Korean sovereign-AI labs. Peer East-Asian-government-aligned AI organizations.
• National research institutes in Thailand, Indonesia, Vietnam, and other Southeast Asian nations. Both partners and operational counterparts in regional language AI development.
• Open-source AI organizations including Hugging Face, EleutherAI, and academic AI labs. Distribution and collaboration partners for SEA-LION releases.

Outlook

Several open questions affect AI Singapore's trajectory in 2026 and 2027:

• The SEA-LION v4 release sequence and the integration of the regionally-aware Vision-Language models.
• The pace of regional adoption of SEA-LION across Southeast Asian governments, enterprises, and developers.
• Continued Singapore government funding commitment and any expansion of the AI programme scope.
• Collaboration trajectory with frontier labs (Google DeepMind, Meta AI / FAIR) and with regional research institutions.
• Talent-development programme outcomes through AIAP and the broader Singapore AI ecosystem.
• The potential expansion of SEA-LION beyond text and embedding modalities into video, audio, and full multimodal capability.
• The programme's role in any Southeast Asian regional AI policy coordination or sovereign-AI initiatives at the ASEAN level.

Sources

• SEA-LION official site. Models, capabilities, and Southeast-Asian-language coverage.
• GitHub: aisingapore/sealion. Open-weights model repository and code.
• Google DeepMind: AI Singapore makes AI more inclusive for Southeast Asia with Gemma 2. Collaboration context.
• Computer Weekly: Sea-Lion explained: Southeast Asia's first large language model. Regional context and capability overview.
• GovInsider: What Singapore's SEA-LION teaches us about the makings of local-language AI. Strategic-policy analysis.
• arXiv: SEA-LION: Southeast Asian Languages in One Network. Research paper.
• AI Singapore: The Team. Programme leadership reference.