GoTo-AI: State-of-the-Art LLMs for the Indonesian Archipelago

GoTo-AI is an open, multilingual AI initiative focused on building high-performance language models tailored for Indonesia and its diverse linguistic landscape. Our mission is to bridge the gap between global large language model (LLM) performance and regional nuance through rigorous continued pre-training and alignment.
We believe in #EmpoweringProgress by making high-performance AI accessible to the Indonesian developer ecosystem. We invite the community to explore, fine-tune, and deploy these models to build the next generation of localized intelligent systems.

Model Suite Overview

The GoTo-AI portfolio is designed for scalability, ranging from edge-compatible 8B models to our flagship 70B model.

• Flagship: 70B Family
  • Our premier instruct-tuned model designed for high-quality generation.
  • Capabilities: State of the art performance in Bahasa Indonesia and English, with deep support for regional languages: Javanese, Sundanese, Balinese, and Batak.
• Scalable Efficiency: 9B & 8B Families
  • Available in both Base and Instruct variants.
  • Optimized for downstream fine-tuning, RAG (Retrieval-Augmented Generation) pipelines, and low-latency production environments.

Technical Foundations

Our models are built using a specialized pipeline to ensure linguistic inclusivity without sacrificing general reasoning capabilities:

• Base Foundations: Leveraging SEALION-based foundations for robust Southeast Asian linguistic priors.
• Training Methodology: Continued pre-training (CPT) on massive, curated Indonesian-centric datasets followed by advanced alignment techniques (SFT/DPO).
• Versatility: Engineered specifically for real-world Indonesian applications, including localized virtual assistants and nuanced conversational AI.