Introducing Raven
Two frontier models, built and served in Canada. Raven Flash and Raven Max are available today.
Today DipoleML launches the Raven model family: two frontier models, Raven Flash and Raven Max, built and served from Canada. They are available immediately in DCode, through the Raven API, and in the Raven web chat at chat.dipoleml.com.
Two models, one family
Raven Max is our deep model. It is built for work where precision carries the most weight: long-horizon reasoning, careful analysis, and decisions you have to defend. Raven Flash is the fast one: frontier capability tuned for production speed, responsive agents, and code at scale.
Both models work across text and images, call tools as they work, and stream their reasoning so you can follow every step. In the Raven app the thinking appears as a collapsible block you can inspect, expand, or ignore.
Where to use them
- DCode, our agentic coding environment, ships with both models today. Download it for Windows, with macOS following within days.
- The Raven API is OpenAI-compatible and pay-as-you-go with prepaid credits. Point your existing stack at it and go.
- The Raven web chat is rolling out now at chat.dipoleml.com. Native apps for macOS, Windows, iPhone, and Android are on the way.
A northern address for frontier intelligence
Frontier intelligence should have a northern address. Raven is engineered, served, and billed from Canada, with a pricing model that starts at zero: a real free tier, subscriptions for heavy use, and pay-as-you-go credits for the API. See the pricing.
This is day one. The family grows from here.
Get started
Try what we just shipped.
The real models, free in your browser, or download DCode and put them to work.