trio.ai is an open-source family of six large language models from RioCloud Solutions, designed to run locally on consumer hardware. The family spans from trio-nano (1.0 GB) — runs on any laptop — to trio-pro (18.6 GB) — frontier-level performance.
pip install triobot — the Python package gives you a CLI plus a Python API. Model weights download as GGUF files from HuggingFace and run via llama.cpp, Ollama or any compatible runtime.
Six models: trio-nano (1.0 GB), trio-small (2.5 GB), trio-medium (4.7 GB), trio-high (5.3 GB), trio-max (7.0 GB) and trio-pro (18.6 GB). All available as GGUF Q4_K_M quantisation on HuggingFace under trioai-org.
Three reasons: privacy (data never leaves your machine), cost (zero per-token charges after install), and reliability (no API rate limits or downtime). Cloud APIs still win on raw capability for frontier tasks; trio.ai wins for edge, on-prem and high-volume workloads.
Yes. trio.ai models are released under MIT licence on HuggingFace, and the triobot Python package is open source on GitHub. Use them in commercial products without licence fees.
Yes. Our trio.ai integration service handles model selection, fine-tuning, quantisation, deployment (edge, server or hybrid) and integration with your existing stack. Book a free consultation.
Read the launch story: Introducing trio.ai — Open Source AI Models for Everyone.
