Open Source • Actively Building

Intelligence
Built Open.
Shared Freely.

Sakitsuna AI is an independent AI research organization crafting open-source machine learning models, innovative AI systems, and cutting-edge tech — freely available for everyone.

Explore Models

Scroll

Our Mission

The Sakitsuna Ecosystem

We believe AI should be open, accessible, and community-driven. Every model we build, every system we architect — shared with the world, no strings attached.

Open-Source AI Models

We develop and release state-of-the-art machine learning models — from language to vision — fully open-sourced on Hugging Face for the community to use, fine-tune, and build upon.

Cutting-Edge Research

From transformer architectures to novel training paradigms, our research explores the frontier of what AI can do — always with a commitment to reproducibility and transparency.

Full Fine-Tuning

We train our language models end-to-end using full fine-tuning — no shortcuts like LoRA or QLoRA. Every weight gets updated to squeeze out the best possible performance.

Community-Centered

We grow with our community. Join our social channels, star our repos, contribute to our projects — every interaction shapes the direction of Sakitsuna AI.

Fast Iteration

We move quickly and ship often. New model releases, experiments, and tooling updates roll out continuously — always pushing the boundary of what's possible.

Responsible AI

Safety and ethics are not afterthoughts. Every model we release is evaluated for potential harms, bias, and misuse — because open AI must also be responsible AI.

Featured Work

Models & Repositories

A selection of our open-source AI projects available on Hugging Face and GitHub. Explore, use, and contribute.

🤗

Sakitsuna Language Models

Open-weight language models optimized for instruction following, creative tasks, and multi-turn dialogue — crafted for real-world deployment.

NLP LLM Open Weights

💬

Instruction-Tuned Models

Fully fine-tuned language models trained to follow instructions precisely — built for chat, Q&A, and task completion with high accuracy and natural fluency.

Instruct Chat Full Fine-Tune

🧠