Router de IA
Router de IA
A single OpenAI-compatible endpoint that reaches many AI models with one API key.
No account with each provider, no separate bill for each one, no code to rewrite when you change model. You add credit, you create a key, and you use it from whatever you already build with.
See it in under a minute
How it works
Four steps, and no subscription anywhere in them.
-
Add credit
Choose an amount and pay the invoice. The credit lands in your account and stays there until you use it.
-
Create an API key
From your panel, in seconds. You choose which models each key may reach, and you can have as many keys as you need.
-
Point your client at the endpoint
Any OpenAI-compatible library, tool or agent: set the base URL and your key, name a model, and it answers.
-
Pay only for what you use
Consumption is discounted from your credit by tokens used. Nothing renews on its own: you top up when you want to.
Where it runs
Three tiers, one endpoint and one key. You choose per request, by naming the model.
Shared local infrastructure, located entirely in Canada and outside United States law.
Dongee's own hardware, not shared with anyone. It does not pass through any datacenter in the United States or in China, and it is not subject to their laws.
Third-party providers, reached through the same endpoint and the same key.
Models and prices
Show the list
Hide the list
You pay for what you use and nothing else. No subscription, no minimum, no charge for keys or requests that fail.
General purpose USD per million tokens
local/gpt-oss-20b
local/llama-3.3-70b
local/mistral-7b
local/mistral-nemo
local/mistral-small-3.2-24b
local/qwen3.5-397b
local/qwen3.5-9b
sovereign/qwen3-30b (coming soon)
nvidia/gpt-oss-20b
nvidia/ising-calibration-1.5-31b
nvidia/laguna-xs-2.1
nvidia/muse-glimmer-30b
nvidia/nemotron-3-embed-1b
nvidia/nemotron-3-nano-omni-30b-a3b-reasoning
nvidia/nemotron-3-super-120b-a12b
nvidia/nemotron-parse-2.0
Programming USD per million tokens
local/qwen3-coder-30b
Image understanding USD per million tokens
local/qwen2.5-vl-72b
local/qwen3.6-27b
local/qwen3.8-27b
nvidia/llama-nemotron-embed-vl-1b-v2
Image generation per image
local/sdxl-base-v10
nvidia/diffusiongemma-26b-a4b-it
Embeddings and search USD per million tokens
local/bge-m3
local/bge-multilingual-gemma2
local/qwen3-embedding-8b
Moderation USD per million tokens
local/qwen3guard-gen-0.6b
local/qwen3guard-gen-8b
nvidia/llama-3.1-nemotron-safety-guard-8b-v3
Embeddings are billed on input only, because they return vectors and not tokens. Image models are billed per image. Speech to text is billed per minute of audio. Everything else is billed on the tokens that go in and the tokens that come out.
Start with whatever amount you want
Sign in to your account to get it. If you do not have one yet, creating it takes a minute.
