GoTo releases open 27B-parameter LLM for Indonesian languages

⚡ The gist in 5 seconds - GoTo has released an open language model with 27 billion parameters focused on Indonesian languages. - The release continues the Sahabat-AI line, launched in 2024 together with Indosat Ooredoo Hutchison. - The architecture, dataset, and license terms are not disclosed in the announcement — only the scale and openness are stated. ### 🔍 What was found Indonesian company GoTo has released an open language model with 27 billion parameters, oriented toward Indonesian languages. According to the release description, this is a continuation of work begun in 2024: back then, GoTo together with telecom operator Indosat Ooredoo Hutchison launched the Sahabat-AI project as part of a digital sovereignty initiative. The source provides no other technical details: there is no model name, no data on context size, training corpus volume, benchmarks, or license. The only concrete figure is the parameter count — 27B, which places the release in the same class as large open models that are usually distributed with their weights, but the announcement contains no confirmation of this. ### 💡 Why it matters GoTo is moving in line with regional "sovereign" LLMs — models tailored to local languages that global multilingual solutions support noticeably worse than Indonesian. The partnership with a telecom operator through Sahabat-AI points to a bet not only on research but also on applied use cases within the country. How competitive the 27B version is on benchmarks cannot be judged from the published description — the actual weights and measurements are needed.