Model settings

Two models, one picker, a thinking level, and 1M tokens of context.

DCode runs on the same two models as everything else. You pick the model per task, and both carry the same capabilities: text and image understanding, tool use, and a streaming reasoning channel.

#The two models

Model	Best for	Context / output
Raven Flash	Everyday coding, quick edits, responsive agent loops, high-volume work	1M context / 128K output
Raven Max	Architecture decisions, hard debugging, long-horizon refactors, anything where being wrong is expensive	1M context / 128K output

Both models accept images: paste a screenshot, drop in a design, or send a photo of a whiteboard, and the model reads it alongside your text.

#Thinking level

DCode exposes the reasoning effort as a thinking-level setting with two values, low and high, applied per conversation. Low keeps quick tasks quick. High gives Max room to plan through the hard ones. The thinking itself renders as an inspectable block: collapsed while you work, one click to read.

#Switching models

Switch from the model picker at any time, mid-project included. Context travels with the conversation, so a common rhythm is: explore and draft with Flash, then switch to Max for the change that has to be right.