Singapore launches upgraded version of Southeast Asia’s LLM MERaLiON

Singapore launched an upgraded version of its large language model (LLM), MERaLiON (Multimodal Empathetic Reasoning and Learning in One Network), on Wednesday. 

Announcing this at the ATxSummit, Singapore’s Minister for Digital Development and Information, Josephine Teo, said MERaLiOn v2 has expanded its language coverage. 

Initially equipped with English, Mandarin and Singlish, the upgraded version would now include Malay, Vietnamese, Thai, Tamil, Bahasa Indonesia, as well as Chinese dialects for future releases.

Marking a key milestone in advancing artificial intelligence (AI), the model also now understands how people in Southeast Asia speak, emote, and interact.  

This is achieved through code-switching capabilities to handle language mixing, which reflects the way people naturally communicate across the region. 

Additionally, the model is also emotionally intelligent, having the ability to detect emotional tone, gender, and paralinguistic features in speech to enable more nuanced, empathetic interactions. 

"This makes MERaLiON relevant to about 450 million people who use these languages primarily on a day-to-day basis," said Minister Teo. 

The Minister added that to ensure that MERaLiON could make a bigger impact, the government will establish a Consortium.  

“A*STAR will partner companies such as DBS Bank, Grab, ST Engineering, NCS, SPH Media Trust, as well as the Ministry of Health to harness expertise in the ecosystem, share learnings, and accelerate adoption,” she added. 

To subscribe to the GovInsider bulletin, click here. 

Three main lines of effort in AI 

Giving an overall perspective on the Singapore’s roadmap on AI, Minister Teo noted that there are three main lines of effort “that are producing good returns”. 

One of them is broad-based access and skills training with around 50,000 or one-third of public officers using a “secure, in-house version of ChatGPT monthly”. 

The second major line of effort involved strengthening core AI expertise in technical government agencies, she added. 

The third major line was to actively transform parts of the public service through AI. In this context, she mentioned the strides being taken in AI for homeland security and healthcare. 

Apart from MERaLiON, there were a slew of other announcements during the first day of the ATxSummit: 

1. Accreditation for good data governance 

From June 2 onwards, organisations will be able to apply for a new Global Cross-Border Privacy Rules (CBPR) Certification, which showcases compliance with internationally recognised data protection standards for organisations. 

The certification was launched by the Global CBPR Forum and will allow certified organisations to transfer data seamlessly across countries within the Global CBPR system. 

Building on the Asia-Pacific Economic Cooperation (APEC), the CBPR system provides certified companies with access to nine economies with about 40 trillion in market size and counting. 

2. Advancing international efforts to enhance AI trust and reliability 
 

ATxSummit also housed a closed-door Ministerial Roundtable hosted by Minister Teo.  

The dialogue between ministers from 18 countries was anchored upon the contents of “The Singapore Consensus on Global AI Safety Research Priorities” document. 

This document is jointly developed by over 100 international AI industry experts as well as government representatives at the 2025 Singapore Conference on AI: International Scientific Exchange on AI Safety last month. 

The living document covers three buckets of research domains across the AI lifecycle – risk assessment, development of trustworthy AI systems, and post-deployment monitoring and intervention. 

It was also aimed at advancing the science of AI safety by identifying research priorities to build a trusted AI system.  

Minister Teo and French Minister Delegate of Artificial Intelligence and Digital Technologues, Clara Chappaz, also signed a joint statement at ATxSummit to collaborate on AI Safety between the two countries. 

The statement outlines Singapore and France’s intent to collaborate on AI Safety, as well as current collaborations around AI Safety Institutes which included domains of knowledge exchanges, governance frameworks, and technical tools. 

3.    Developing an AI talent pipeline 

IMDA also announced several initiatives on Tuesday, which were aimed at equipping enterprises to strengthen cyber resilience with advanced AI capabilities and quicken the adoption of frontier technologies. 

One of these was partnerships with private sector companies. 

IMDA would partner with Alibaba, Prudential Singapore and ST Engineering respectively to enable SMEs to build their competencies in cloud, AI, generative AI (Gen AI), and cybersecurity. 

For digitally mature SMEs (DMEs), IMDA would work with Amazon Web Services (AWS) and Microsoft through its GenAI x Digital Leaders initiative launched in 2024 to accelerate GenAI adoption.  

Over 1,000 enterprises and 500 projects would be supported by the expanded initiative that features tech discovery and deep-dive sessions, as well as one-on-one consultation opportunities. 

Moreover, IMDA would collaborate with AI Singapore (AISG) and over 25 leading companies to provide 800 AI practitioner job and training opportunities for individuals. 

Half of these opportunities will be open for applications through IMDA’s TechSkills Accelerator that provides jobs and training opportunities with companies like AWS, Microsoft, Singtel, and Oracle. 

IMDA’s partnership with AISG also saw the launch of two new training programmes, the AI Apprenticeship Programme (Industry) (AIAP(I)), and Pinnacle AI Industry Programme (PAIP), for the other 400 opportunities.