Models & routing
Auto router
Send minirouter/auto and MiniRouter picks the model.
Send a requestExamples read your key from MINIROUTER_KEY.
curl https://api.minirouter.sh/v1/chat/completions \
-H "Authorization: Bearer $MINIROUTER_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "minirouter/auto",
"messages": [
{
"role": "user",
"content": "Explain an API gateway in one sentence."
}
],
"max_tokens": 256
}'Intents
- minirouter/
auto :smart, or:codewhen the request has tools or JSON output.- minirouter/
auto: smart - A strong general model at a mid-range price.
- minirouter/
auto: cheap - The cheapest model that still answers well.
- minirouter/
auto: code - Reliable tool calling and long context.
- minirouter/
auto: roleplay - Long context with low input cost.
- minirouter/
auto: free - Free models only. See Auto Free.
To choose from what the request says, send minirouter/fusion.
Every router has its own page, starting with Auto.
How it choosesLater models in the list become fallbacks. Auto never reads prompt text.
- AllowedOn your key.
- AffordableWithin your cost tier.
- HealthyHas a working provider.
- FitsCan accept this request.
The first model to pass answers. See the live lists at Auto models.
Cost tierOutput price ceiling per million tokens. On Low, bare minirouter/auto starts from :cheap.
- Low
- $1
- Medium
- $5
- High, the default
- Any
Set it in Routing settings.
Response
- Chosen modelNamed in
modelandx-minirouter-model. - PriceProvider cost plus 8%.
- No match503
auto_unresolvable. Retry after 15 seconds.
Limits
- Endpoints
- Chat Completions, Responses and Messages.
- Fallback lists
- Not combinable with
modelsorfallbacks.