Where we are
Roadmap
Where Dawnstack is today and what we are building next. There are no dates here on purpose: a date is a promise, and we would rather tell you what is true now.
Shipped
-
Dawnstack opens
Chat completions on over 30 models, reached by changing one base URL.
-
API keys and per-request usage
-
Top-ups in local currency
Paid the way you already pay, no card required.
-
Streaming responses
-
Cost on every response
Returned with the request, not reconstructed later, on streamed replies as well as whole ones.
-
Language cost measured and published
Eight African languages for text and nineteen for speech, per model, with the ones we have not measured marked as not measured.
-
Tool calling and structured output
Measured per model with real requests, because the suppliers' own capability tags are wrong in places.
-
Prices in your own currency
Detected from where you are, and switchable.
-
Speech, in and out
Transcription and synthesis on the same base URL, priced per second of audio and per character of text.
-
A starter balance on every account
Enough for a few hundred requests on the cheaper models, so you can tell whether this works before you pay for it.
-
The playground
Try a model before you write any code, billed the same as your own client.
-
Connect an AI assistant
Claude, ChatGPT and Cursor can read the catalogue and these docs, then build against the API. Connecting is a click.
In flight
-
Vision
Images are accepted by 8 of the models we serve. Sending one works; the cost estimate we reserve against does not account for it yet.
-
More measured languages
They appear on the models page when they are measured, not when they are planned.
Ahead, in this order
-
Embeddings
So retrieval over your own documents does not send you back to a provider that wants a dollar card.
-
Dawn Studio
A building surface for people who do not write code. Today Dawnstack is an API, so you need to write code to use it.
Something you need that is not on this page? Tell us at support@dawnstackai.com.