Africa-first AI infrastructure
Frontier AI, priced in your currency
Paid how you already pay. Point your existing OpenAI code at Dawnstack, change one URL, and keep everything else. No dollar card, no monthly minimum.
Top up from$1.001.00 US dollars, shown converted to local currencies at today's rate.
All 42 models we serve are open-weight today. You can download the weights, run them yourself, and check our measurements.
The constraint
Affording it and being able to pay are not the same problem
You cannot pay
Most providers start at five dollars and want a dollar-denominated card. Without one, no amount of willingness opens an account.
The bill ignores your language
A flat rate per million tokens assumes a token is worth the same in every language. In Swahili the same sentence takes about twice as many.
The mechanism
One base URL, and the response your client already parses
curl https://api.dawnstackai.com/v1/chat/completions \ -H "Authorization: Bearer $DAWNSTACK_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "llama-3.1-8b", "messages": [{"role": "user", "content": "What is the capital of Kenya? One sentence."}] }'
{
"id": "chatcmpl-744d0449-6366-44c7-96e7-a8a1f43ac520",
"object": "chat.completion",
"created": 1785097284,
"model": "llama-3.1-8b",
"choices": [
{
"index": 0,
"message": { "role": "assistant", "content": "The capital of Kenya is Nairobi." },
"finish_reason": "stop"
}
],
"usage": { "prompt_tokens": 21, "completion_tokens": 8, "total_tokens": 29 },
"dawnstack": { "cost_micros": 6, "cost_usd": 0.000006, "request_id": "req_8f2a1c" }
}
Base URL https://api.dawnstackai.com/v1.
The response is the OpenAI shape, so your existing client parses it unchanged.
How it works
Four steps, and only one of them touches your code
STEP 1
Sign up
Email or Google. No card, and a starter balance to try it.
STEP 2
Create a key
One click. Shown once, stored as a hash.
STEP 3
Change one URL
Point your OpenAI client at us. Nothing else changes.
STEP 4
Top up
When the starter balance runs out. In your own currency, whenever you need it.
Every account starts with a small balance, enough for a few hundred requests on the cheaper models. After that your balance pays for requests: top up in your own currency, or redeem a code if you have one.
Measured, not claimed
We publish what your language actually costs
A price per million tokens assumes a token is worth the same in every language. It is not. The same sentence in Swahili takes about twice as many tokens as in English, and in Amharic several times more, so identical meaning costs more. A single flat rate passes that on without showing it.
We measured it on FLORES-200, 1,012 professionally translated sentences, against every model we serve, and publish the multiplier per model and per language. Where a model is licensed so that nobody can serve it commercially, we say that too.
Models
over 40 models, one endpoint
Models from Meta, Google, OpenAI, Mistral, Alibaba and Moonshot. Switch between them by
changing the model
field. Prices are per million tokens.
mistral-nemo
Mistral AI
$0.03
per 1M in
$0.04
per 1M out
qwen3-vl-30b
Qwen (Alibaba)
$0.23
per 1M in
$0.90
per 1M out
mimo-v2.5
Xiaomi
$0.60
per 1M in
$3.00
per 1M out
hermes-3-llama-3.1-70b
Nous Research
$1.05
per 1M in
$1.05
per 1M out
gemma-4-e4b
$0.03
per 1M in
$0.15
per 1M out
For your agent
Your coding agent can read the catalogue and write the code
Connect Dawnstack to Claude, ChatGPT or Cursor and your assistant reads the live prices, the measured language costs and the reference, then builds against the API. Connecting is a click, with no key to paste.
Paste it into Claude, ChatGPT or your editor and it connects itself.
Or add it yourself
https://api.dawnstackai.com/mcpClaude
ChatGPT
Claude Code
Cursor
Gemini CLI
Anything that speaks the Model Context Protocol. Setup differs by app.
Pricing
Your first top-up can be one dollar
No subscription, no card on file, and your balance never expires. You are billed for the tokens a request actually produces, and your balance is denominated in US dollars.
No subscription
Nothing recurring, and nothing to cancel.
No expiry
Your balance stays on the account until you spend it.
No card on file
Nothing is stored to charge you later.
No minimum
Top up what you need, and come back for more when you need it.
$1.00 is roughly 22.2 million output tokens on mistral-nemo. See what a request costs.
What is built, and what is next
Chat completions, tool calling, speech, top-ups in your own currency and measured language costs all work today. Embeddings and a building surface for people who do not write code are what we are working on.