CrowdRouter
Join beta

Docs

Documentation

The API is OpenAI-compatible. Point an existing client at our base URL and change the model ID.

Not yet serving. The endpoint is documented and the models are benchmarked, but no capacity is in service. Access opens through the beta first.

Base URL

https://api.crowdrouter.com/v1

Client setup

Any OpenAI-compatible client works. No custom SDK is required.

python
from openai import OpenAI

client = OpenAI(
    base_url="https://api.crowdrouter.com/v1",
    api_key=os.environ["CROWDROUTER_API_KEY"],
)

Supported request fields

We support the fields below and reject anything else with a clear error rather than silently dropping it.

  • /v1/chat/completions
  • streaming
  • stream_options.include_usage
  • max_tokens
  • temperature
  • top_p
  • stop
  • seed
  • n
  • tools / tool_choice
  • reasoning_effort
  • response_format

Model IDs

Model lifecycle

Models are added when they meet the bar below, and retired through a published process: announced, dated for removal, and any active usage is notified before the date.

How models are selected

We don't try to host everything. A model earns a place in the catalog by clearing every one of these:

  • Real demand from developers running recurring workloads
  • Useful capability for the workloads people actually run
  • Serving efficiency on commodity hardware
  • Licensing we can serve commercially
  • Measured, reproducible performance
  • Pricing that stays profitable as inference economics move

About CrowdRouter

CrowdRouter is an affordable inference API for useful open models. The catalog changes; the product does not.

Join beta