Intelligence
changes poles.
Canada’s frontier model company. Proprietary multimodal intelligence, trained domestically and built to move the centre of gravity north.
Raven 1.0 · Arriving 01.10.2026
Two models.
One new frontier.
Raven is trained across language, vision, audio and code as one continuous system. It reasons through work, not just prompts.
Raven 1.0 Mini
Deep intelligence
High-order reasoning for research, analysis and decisions where precision carries the most weight.
Try the previewRaven 1.0 Flash
Intelligence at velocity
Frontier capability engineered for production speed, responsive agents and code at scale.
Try the previewPreview evaluation
Measured against the frontier.
SWE-bench Verified performance. Raven figures are internal pre-release projections, not independently validated. Competitor figures are public reported scores.
Public figures: OpenAI GPT-5 system card, Anthropic Claude Opus 4.1 announcement, Google Gemini 2.5 Pro model card. Settings and harnesses differ, so scores are directional rather than a controlled head-to-head result.
The Canadian thesis
Sovereign by design. Global by ambition.
Our models are trained in Canada, run through Canadian infrastructure and developed by a Canadian team. This is more than data residency. It is a national capability designed to compete at the highest level.
Why Canada, why nowDCode
From intent to working software.
A native environment for teams that build with agents. Plan across repositories, execute in parallel, inspect every change and ship without leaving the flow.