Choose your balance with Fusion
Fusion reads your request and chooses a model, with automatic reasoning effort where supported. Now you can choose how strongly it favours price or quality, without maintaining your own model list.
Three profiles. One API.
Fusion Cheap
Spend lessA stronger preference for lower estimated cost.
Fusion
Find a balanceThe default balance of benchmark quality and estimated cost.
Fusion Max
Aim higherHigher quality targets, with more room for cost.
A frontier for each task
Fusion compares models and reasoning efforts on benchmark quality, estimated request cost and latency. It sorts requests into coding, reasoning, writing or conversation, with four complexity levels for each.
Each profile has its own picks and fallbacks across those 16 combinations. The picks come from published settings, so they can improve as models, prices and evaluations change.
Benchmarks guide selection; they do not guarantee a better, cheaper or faster answer to every prompt.
Switch with one model ID
Use the profile you want on Chat Completions, Responses or Messages. Messages keeps the model’s default reasoning effort. Existing calls to keep the balanced profile.
curl https://api.minirouter.sh/v1/chat/completions \
-H "Authorization: Bearer $MINIROUTER_API_KEY" \
-H 'Content-Type: application/json' \
--data '{
"model": "minirouter/fusion:cheap",
"max_tokens": 1024,
"messages": [{ "role": "user", "content": "Fix this failing test." }]
}'Your explicit reasoning setting still wins. The response header names the model that served the request; shows the task and tier when classification succeeds.