Sahabat-AI
Indonesia's sovereign open-source LLM family covering Bahasa, Javanese, Sundanese, Balinese, and Bataknese
Sahabat-AI is an open-source Indonesian LLM platform best for SMEs, banks, and public-sector teams that need Bahasa-first AI under PDP Law data residency rules. Its SEA edge is sovereign infrastructure — 70B open weights running on Lintasarta's GPU Merdeka inside Indonesia, plus native coverage of Javanese, Sundanese, Balinese, and Bataknese alongside Bahasa and English. The caveat: at 38 reviews and counting, the production ecosystem (tooling, fine-tuning shops, monitoring) is still nascent compared to GPT-4 or Claude, so deployment work is more DIY than buying OpenAI access.
- ✓Free, open-weights 70B model under permissive license for self-hosting
- ✓Five Indonesian local languages including Javanese and Sundanese — unmatched coverage
- ✓Hosted on Lintasarta GPU Merdeka inside Indonesia — clean PDP Law data residency story
- ✓Smaller 8B and 9B variants reduce GPU cost for SMEs
- ✓Backed by GoTo and Indosat — credible long-term funding and distribution
- ×Production tooling and ops ecosystem still thin compared to GPT-4 or Claude
- ×Self-hosting requires meaningful MLOps capability that most Indonesian SMEs lack
- ×Reviews count of 38 reflects how early the production deployments are
- ×Outside Indonesia, no native Thai, Vietnamese, or Tagalog coverage
About Sahabat-AI
Sahabat-AI is the open-source Indonesian LLM project led by GoTo and Indosat Ooredoo Hutchison, with infrastructure backed by Lintasarta's GPU Merdeka. The flagship model is now 70 billion parameters with a multilingual chat service at sahabat-ai.com and inside the GoPay app. Models are downloadable from Hugging Face for self-hosting under permissive licenses.
Key Features
Best For
Southeast Asia Fit
Sahabat-AI is Indonesia's clearest answer to the sovereign-AI question. The infrastructure runs on Lintasarta's GPU Merdeka inside Indonesia, the model speaks five regional languages spoken by hundreds of millions of Indonesians, and the open weights mean any team subject to PDP Law data residency rules can self-host without sending prompts overseas.
- Global
- SEA
Integrations
- Payment
- GoPay
- Other
- Hugging Face
- GPU Merdeka
- Lintasarta
- vLLM
- GoPay app
- OpenAI-compatible API
"Verified" means read on the vendor's own published pages on the date shown. We do not run hands-on tests, and a statement a vendor confirmed to us directly is labelled as vendor-confirmed. Software-listing.com is independent and may earn affiliate commissions from some links.
Related Analysis & Guides
The questions operators actually ask.
Is Sahabat-AI free to use?
Yes. Models are downloadable from Hugging Face under permissive licenses, and the hosted chat at sahabat-ai.com is free. Self-hosting cost depends on your GPU infrastructure.
Does Sahabat-AI satisfy Indonesian PDP Law data residency?
Yes, if self-hosted or used via the GPU Merdeka infrastructure inside Indonesia. The whole point of the project is sovereign AI that keeps prompts and data inside the country.
Can Sahabat-AI handle Javanese and Sundanese?
Yes, those are first-class supported languages alongside Bahasa Indonesia, Balinese, Bataknese, and English. This is the model's clearest moat versus GPT-4 or Llama.
Is Sahabat-AI as capable as GPT-4 for Bahasa Indonesia?
It depends on the task. For Bahasa and regional languages, the 70B model is competitive. For complex reasoning and English-heavy tasks, GPT-4 still wins. Most production teams use both.