MERaLiON
Singapore's national multimodal speech-text LLM built for Southeast Asian languages, accents and Singlish code-switching.
Overview
MERaLiON (Multimodal Empathetic Reasoning and Learning in One Network) is Southeast Asia's speech-text large language model, developed by A*STAR's Institute for Infocomm Research under Singapore's S$70M National Multimodal LLM Programme, backed by the NRF and IMDA. First released in December 2024, its later versions cover English, Mandarin, Malay, Tamil, Singlish, Bahasa Indonesia, Thai and Vietnamese, handle code-switching, and add audio understanding and emotion detection. Model weights are published openly on Hugging Face.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 24 Jul 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
ASEAN Perspective
MERaLiON in Southeast Asia
ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).
What this is for: A regionally-tuned speech-text LLM that understands Southeast Asian languages, accents, Singlish and code-switching, with audio and emotion understanding.
Who this is for: Researchers, developers and Singapore/SEA enterprises building multilingual voice and language applications for local users.
Availability: Free open-weights on Hugging Face; a government-backed research programme with an industry consortium rather than a hosted consumer product.
What people say
Don't look for MERaLiON on G2 or an app store; it's a research programme, not a product. Its reputation is built on Hugging Face downloads and academic papers, like the ACL 2025 demo acceptance for its AudioLLM work. The whole model family—AudioLLM, MERaLiON-2-10B, and the newer MERaLiON-3-10B—is published openly. That peer-reviewed acceptance is meaningful signal for a national research programme rather than a marketed app.
On benchmarks the team positions MERaLiON-3-10B against Qwen3-Omni, Gemini Flash, GPT-4o Audio and its own predecessor, reporting best-in-class results on roughly 31 of 59 tasks spanning ASR, spoken QA, emotion and speaker-attribute recognition. Practitioners point to its Southeast Asian language coverage—Singlish, Malay, Tamil, plus Hokkien and Cantonese—as the main draw. It also offers end-to-end audio reasoning, not just a transcribe-then-prompt pipeline.
The caveats are straightforward. Winning on about half its own benchmarks means it still trails frontier models on many tasks. And since the programme authors most evaluations, independent comparisons are thin on the ground. As a 10B open-weights model, it needs GPU hosting and real engineering to deploy. The emotion-inference claims, in particular, should be met with skepticism until outside groups can reproduce them.
Summary of public user & expert reviews, compiled by RECATOOLS.
About this listing
This entry was compiled from publicly available data including MERaLiON's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with MERaLiON unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to MERaLiON directly →
Spotted something out of date? Suggest an update →
Alternatives to MERaLiON
More in LLMs & Chat
