- License
- Apache 2.0
- Model ID
- qwen3.8-27b
Affordable inference for useful open models.
An OpenAI-compatible API for open models, without GPUs to run or a minimum spend. The beta launches with Qwen3.8-27B.
$0.20 input · $0.05 cached input ·$1.10 output per 1M tokens · 128K context
Beta access opens gradually. Beta pricing is subject to change during the beta period.
from openai import OpenAI
client = OpenAI(
# CrowdRouter's address instead of OpenAI's
base_url="https://api.crowdrouter.com/v1",
api_key=os.environ["CROWDROUTER_API_KEY"],
)
reply = client.chat.completions.create(
model="qwen3.8-27b", # our model
messages=[
{"role": "user", "content": "Hello"},
],
)Built to keep your costs down
Straightforward usage-based beta pricing
Pay per token, with new input, repeated input and output priced separately. Beta pricing is subject to change during the beta period.
Nothing to pay when you are idle
You pay per token. There is no minimum spend, no subscription and no GPU to rent, so a quiet month costs nothing.
Repeated context has its own rate
Chats and agents send the same conversation again on every turn. We bill that repeated context separately from new input, so you can see what it costs.
$0.20 input, $0.05 cached input, $1.10 output, per million tokens
Beta pricing is subject to change during the beta period.
Models
We are starting with one model and adding more over time.
We add a model when we can serve it at an affordable price. All models →
Up and running in three steps
Join the beta
Tell us what you plan to run. It takes a minute.
Get your API key
We'll email you as soon as your access is ready.
Change one line
Point your OpenAI client at CrowdRouter and pick a model.
Questions
How much does it cost?
You pay per token, with input, cached input and output priced separately and no minimum spend. Input is $0.20, cached input $0.05 and output $1.10 per million tokens during the beta.
Is CrowdRouter live?
Not yet. We're opening access gradually through a private beta. Join the list and we'll email you when your key is ready.
Which models can I use?
The beta starts with Qwen3.8-27B, a mid-size open model (Apache 2.0 license). We add more when we can serve them at an affordable price.
Will my existing code work?
If it already calls the OpenAI API, yes. You change two things: the address it sends requests to, and the model name. Streaming, tool calls and structured output are supported.