All The Models
Choose models, manage keys, and track usage from one clean AI layer.
Live request
Route any model.
See every supported model.
Filter by company, endpoint type, and price tier. Open code samples from the public catalog.
A calmer control layer for every model call.
Choose the model explicitly, keep keys contained, and watch spend move in one soft dashboard.
Choose any model, on purpose.
inference does not hide model choice behind vague automation. Send the exact model ID your request needs, then keep fallbacks, keys, usage, and cost visible in one place.
Ship safer keys
Create scoped keys for apps, teams, and environments while keeping credentials off projects.
Control spend
Prepaid credits, per-key limits, and live cost visibility keep experiments inside the budget you set.
Use more surfaces
Chat, responses, images, embeddings, and new model surfaces stay organized under one dashboard.
All the models, without making your AI layer louder.
One endpoint, explicit model choice, scoped keys, and usage that stays readable as your product grows.
Change the base URL. Keep the workflow.
Use the same SDK shape, pass the model you want, and see routing, cost, and limits beside the request.
from openai import OpenAI
client = OpenAI(
api_key="sk-air-your-key",
base_url="https://inference.ws/v1",
)
response = client.chat.completions.create(
model="gpt-4.1",
messages=[
{"role": "user", "content": "Route this through inference."}
],
)
print(response.choices[0].message.content)Latency
612 ms
Cost
$0.0041
Questions people ask first.
Public pages show the catalog and pricing. Account-specific keys, limits, and usage live in the dashboard.
Yes. The public catalog shows available models, pricing, endpoint type, and code samples.
Yes. Keep the OpenAI SDK shape, set the inference base URL, and pass the model ID in the request.
Create scoped keys for apps, environments, or teammates. Limits and usage are tracked per key.
The dashboard shows credits, rate limits, request history, and model-level costs after you create an account.
Plans that match the real router limits.
Free
Yearly: $0 credit / year
For trying the router and validating basic chat flows.
- Chat models
- 10 RPM
- 15k TPM
- No monthly credits
Starter
Yearly: $48 credit / year
For small apps moving past playground experiments.
- Chat models
- 30 RPM
- 60k TPM
- $5 monthly credits
Pro
Yearly: $240 credit / year
For production apps using chat, responses, and images.
- Chat, responses, image
- 120 RPM
- 250k TPM
- $25 monthly credits
Ultra
Yearly: $960 credit / year
For teams routing heavier traffic across all modalities.
- All modalities
- 600 RPM
- 2M TPM
- $100 monthly credits
One AI layer. Less wiring. More shipping.
Start in the playground, then move the same key into production.