Everything Dune can do, from your own code
The same models, the same reasoning and the same care about sources — through one API, priced per token.
One call, any model, replies as they are written
The API speaks the same way Dune does. Send the input and a model name; get the answer streamed back, with the reasoning and the sources when you ask for them.
- Streaming by default, with a single flag to turn it off
- Files, images and long documents as input
- SDKs for Python, TypeScript and Go
Request
model: "dune-ridge" input: "List the risks in this contract" max: 800 stream: true
Response
The three risks are the auto-renewal clause in §4,
the uncapped liability in §9, and the 90-day
notice period for termination in §12.
412 tokens · 0.9s · streamed
Three models, one shape of answer
Pick by the size of the problem. They share the same training, the same care about sources, and the same API.
| Model | Best for | Context | Speed | Input | Output |
|---|---|---|---|---|---|
|
Sand dune-sand |
High-volume work where speed and price matter: classification, extraction, short replies. |
200K | Fastest | $0.80 | $4 |
|
Ridge dune-ridge |
Everyday reasoning, writing and coding. The default for Pro and for most API traffic. |
500K | Fast | $3 | $15 |
|
Summit dune-summit |
The hardest problems: long research, multi-step agents, difficult code across a whole repository. |
1M | Deliberate | $15 | $75 |
Prices per million tokens. Context windows in tokens.
Works where your team already works
Connect the tools once; Dune reads and writes to them with your permissions and nobody else’s.
Start building this afternoon
Free API credits on every new account. No card until you need more.