SINGAPORE, May 5 — Singapore’s home-grown large language model (LLM), Sea-Lion, has reportedly recorded about 235,000 downloads, driven by adoption from firms including Indonesia’s GoTo Group.
According to a report in The Straits Times the model currently recognises 13 languages, including Malay, Tamil, Thai and Vietnamese, and is set to add voice recognition in 2025, followed by visual capabilities.
GoTo Group was reportedly among the earliest users, having integrated Sea-Lion into its own artificial intelligence (AI) system in early 2024.
“Training a model from scratch is often prohibitively expensive. So like many in the industry, we adopted a continuous pre-training approach.” GoTo’s chief data officer Ofir Shalev was quoted as saying.
Sea-Lion is being developed by AI Singapore, a national research and innovation programme aimed at boosting the country’s artificial intelligence capabilities.
Leslie Teo, lead of the Sea-Lion team, told The Straits Times that the model now performs “close to the frontier” and can serve as a companion to larger models such as ChatGPT.
AI Singapore hopes to expand adoption and gather feedback from more enterprise users.