AI Singapore and Dell Technologies collaborate to democratize AI access in Southeast Asia

The collaboration will see the SEA-LION model made available to Dell Technologies’ AI PCs and edge devices for testing. However, there are no immediate plans to integrate this specific AI model into Dell's devices.

To ensure AI is accessible to everyone across the Southeast Asia region, AI Singapore (AISG) is working with Dell Technologies to enhance its SEA-LION (Southeast Asian Languages in One Network) family of open-source large language models (LLMs). Specifically, the partnership will see both Dell and AI SG testing and validating SEA-LION models across various Dell AI PCs and edge infrastructure.

Given that most AI models prioritize English, there have been concerns that there could be gaps among countries whereby English is not the primary language of communication. In Southeast Asia, the wide diversity in cultures and languages as well as local slangs within the English language can make it challenging in the development and deployment of AI use cases.

SEA-LION bridges this gap by training open-source models on a vast, high-quality corpus of region-specific data. This approach embeds local linguistic characteristics, enabling hyper-local multilingualism and fostering innovation tailored to the region. As an open-source initiative, SEA-LION also democratizes AI by reducing the high costs associated with foundational model training. Its permissive licensing encourages community contributions, supports regional control and provides an accessible alternative to proprietary systems.

For Dell, deploying SEA-LION on AI PCs and edge devices with AI SG not only ensures great access but allows users to have leverage the vendor’s specialized expertise in AI infrastructure optimization.

According to Andy Sim, vice president and managing director, Singapore, Dell Technologies, the collaboration demonstrates that sophisticated, culturally intelligent AI can run efficiently on laptops and edge devices.

“Together, we are democratising AI in Southeast Asia and are fostering a future where advanced AI is accessible to all,” Sim said.

For Dr. Leslie Teo, senior director, AI Products, AI Singapore, the collaboration is a key step in realising the vision for SEA-LION, whereby models that are not only local and culturally relevant but also resource-efficient enough to be deployed on the edge.

“By running fully featured LLMs directly on Dell devices, we provide enterprises with privacy and lower cost alternatives while remaining responsive and reliable,” said Dr Teo.

A deeper look at the collaboration

CRN Asia reached out to Dell Technologies to understand more about the collaboration, especially on how the AI models will be integrated into Dell’s AI PCs as well as the challenges they could face while testing out the models.

Sim shares more with CRN Asia in the replies below.

Apart from testing, are there plans to integrate the AI models into some of Dell's AI PCs?

As an open-source initiative, SEA-LION is free for all to use. Its permissive licensing encourages community contributions and democratises AI by reducing the high costs associated with foundational model training.

As part of our current collaboration with AI Singapore, Dell AI PCs and edge infrastructure are being used to test and validate the SEA-LION models. This supports AISG's efforts to build models that are resource-efficient and can be deployed on lightweight setups. The AI models do not need to be integrated into the AI PCs for testing and validation purposes. While the insights from this testing help inform and validate use cases for our AI PCs, there are no immediate plans to integrate this specific AI model into Dell's devices.

Can you share any use cases that Dell is working on with AI Singapore on this?

Dell Technologies and AI Singapore are partnering to make local AI more accessible and enterprise-ready. By optimizing the SEA-LION language models for Dell AI PCs, we are enabling organizations to operate advanced AI fully offline with greater privacy, lower latency and reduced cloud dependence. This collaboration also brings real-time speech transcription to enterprise workflows and ensures reliable performance on NPU-accelerated devices. Together, we are turning AI potential into practical progress for Southeast Asia.

While SEA-LION is capable of understanding 11 Southeast Asian languages, what will be the biggest challenge in testing and how is Dell dealing with the challenges?

One of the key challenges lies in ensuring consistent accuracy and performance across diverse linguistic and cultural contexts. Although SEA-LION is trained to understand multiple Southeast Asian languages, each language has varying dialects, idiomatic expressions, and contextual nuances that can influence model precision. Testing therefore requires extensive local validation with real-world datasets. Dell is also committed to ensuring responsible AI development, paying particular attention to data privacy, inclusivity, and fair representation.

Dell uses its full range of AI-ready infrastructure, from AI PCs to edge systems, to rigorously test SEA-LION under diverse deployment conditions. This includes evaluating offline performance, hybrid cloud-edge environments, and various hardware configurations. Together with AI Singapore, we use real-world datasets that reflect Southeast Asian linguistic patterns and conversational nuances to ensure the model performs reliably in authentic contexts. By stress-testing SEA-LION against these region-specific scenarios, we ensure it remains efficient, secure, and production-ready for organizations across the globe, no matter how or where it is deployed.

Will this be available only in Singapore or available to other SEA nations as well?

While SEA-LION is being developed and tested in Singapore, the initiative is designed to benefit the broader Southeast Asian region. As an open-source model family, SEA-LION supports collaborative development across regional partners, enabling researchers and organizations in other SEA nations to adapt it for local use cases.