AURORATry it free

No. 11ApolloInference API

One inference API for every open and frontier model — stream tokens in milliseconds, keep a single request schema across the whole model ladder, and ship knowing your data never trains a model.

5M free tokens / monthno card38 ms median first token

Powering inference for

Cohere models served via Aurora marketplace partner. Logos shown with permission.

Live playground · runs in your browser

Type a prompt. Watch it tokenize, then compose.

A faithful mock of the request envelope — no key, no network, no rate limit. Tokenize on the left, stream a grounded response on the right, and feel the 38 ms contract before you ever paste a secret.

Tokens
  • Expl
  • ain
  • atten
  • tion
  • in
  • one
  • line.
Response
Run the prompt to stream a grounded response. Cmd-Enter works too.

Platform · 03 instruments

The rack the aurora is wired into.

Server-sent token streaming over HTTP/2 with a stable JSON contract, so first-token latency is the only number your UI ever waits on.

transport
HTTP/2 · SSE
first token
38 ms p50
contract
stable JSON

Swap a model id and keep your code. Tools, structured output, and vision share a single request envelope from the 1B draft model to the frontier tier.

envelope
/v1/responses
ladder
1B → frontier
shared
tools · json · vision

Zero-retention inference, regional routing, and SOC 2 Type II controls are the default — not an enterprise upsell hidden behind a sales call.

retention
0 days default
routing
EU · US pinned
training
never

No. 05 / Scope · live telemetry

The three numbers we live by.

Read straight off the rack — median time-to-first-token, sustained decode rate, and the uptime we hold every region to. No averages that hide the tail; these are the figures the playground above is wired to.

38ms
median first token
1280tok/s
peak throughput
99.98%
rolling uptime

Trust console

Three answers your security team already wrote down.

ALL CONTROLS NOMINAL · LAST AUDIT 2026-04

Free tier · live now

Your first 5M tokens are on us.

Create a key, paste two lines, and stream your first completion before this aurora finishes a single pass. Usage-based after that — with hard caps you set, never a surprise invoice.

Monthly grant
5,000,000 tokens
Card required
None
Time to first token
38ms
Get an API keyOpen the playground

Self-serve · no sales call · revoke any key in one click