Guardian 1.0 Thinking — Release
Our flagship reasoning model is almost here. Long context, privacy-first architecture, encrypted in transit and at rest.
Guardian 1.0 Thinking is coming soon. This is the first model in the Guardian family and the foundation of everything we are building at OQENYX.
Guardian 1.0 Thinking is a reasoning-optimized language model with particular depth in STEM, law, medicine, and software engineering. It was designed from the start with two constraints that shaped every product decision: privacy-first architecture, and operability with minimal retention.
The model ships with three reasoning modes. Low reasoning mode delivers fast, direct responses for conversational tasks and light analysis. Medium mode activates expanded chain-of-thought processing, suitable for most professional and technical workloads. High mode enables deep multi-step reasoning — ideal for complex research, legal analysis, and advanced code generation. Reasoning depth is selectable per-request.
The full long context window is available across all reasoning modes from day one — no tiered access, no hidden limits.
Privacy architecture: all inference runs in isolated compute environments with no cross-tenant data sharing. Prompts and completions are encrypted in transit and at rest. Minimal retention by default means no logs, no caches, no traces beyond the active request lifecycle. For organizations that need audit trails, optional logging is available with full encryption and user-controlled retention periods.
Guardian 1.0 Thinking will be available through the Onora App at launch. Pricing is per-token with no seat fees. Enterprise contracts with dedicated inference capacity are available by request. Get started at onora.app or contact us for enterprise onboarding.
Guardian Lite, our fast lightweight model, and Guardian Vision, our multimodal model, are in active development. Stay tuned.