OrcaRouter

OrcaRouter is an AI research and infrastructure team building secure, open, and accessible tools for AI inference, model routing, and AI security.

Our work focuses on adaptive model routing, inference acceleration, AI red teaming, blue teaming, and model security evaluation. We develop and release models and research tools designed to help developers and security researchers study model behavior, adversarial robustness, safety mechanisms, and emerging AI security risks.

We also maintain OrcaRouter, a unified AI inference gateway providing access to a broad range of open and frontier models through a single OpenAI-compatible API.

Our Hugging Face releases include research-oriented and uncensored model variants intended for authorized security research, red-team testing, adversarial evaluation, and model behavior research. (Released models are hosted here: https://www.orcarouter.ai/models?q=uncensored)

🌐 OrcaRouter: https://www.orcarouter.ai

📄 Research: https://arxiv.org/abs/2605.30736

💻 Open Source: https://github.com/Continuum-AI-Corp/OrcaRouter-Lite