modeliddecisionnode(todaydecisionnode-main-0.9.1-rc1)- Answers
- Choice, Score, Truth, Number, Point and Box, each calibrated
- Serves
/v1/decide, sessions and batch jobs- Input
- Text, JSON, up to 16 images and video as frames in one request
- Context
- 64k tokens per request (65,536)
- Price
- $0.042 per 1M input tokens, output free
- Images and video
- $0.09 per 1M image and video tokens, no fee per image
- Batch jobs
- 30% off: $0.0294 per 1M input tokens
- Runs on
- Our own inference stack and our own GPUs
- No ready capacity
- Answered at once by DecisionNode Flash, unless you opt out: see below
When to use it#
Use the full model when the decision is subtle: long threads, many options, images with small print, rubrics where the levels are close together. It is the default in these docs and in the playground.
Both models are System One models: one pass, no deliberation, a calibrated probability over the options you allowed (what that means). Pick between them on accuracy, speed and price, not on what they can do; they answer the same question types.
How it answers#
It reads the state, the images and every question in one pass and returns every answer directly. It generates no text, which is why output is free and why the answer always has the shape its type defines.
Calling it#
curl https://api.decisionnode.com/v1/decide \
-H "Authorization: Bearer $DECISIONNODE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "decisionnode",
"state": "Customer: I was charged twice and nobody has replied for 3 days.",
"questions": {
"route": {
"type": "choice",
"instructions": "Where should this go?",
"criteria": {
"billing": "money",
"bug": "broken",
"account": "login"
}
},
"urgency": {
"type": "score",
"instructions": "How urgent is this?",
"criteria": ["routine", "today", "urgent", "critical"]
},
"refund": {
"type": "truth",
"instructions": "Refund this automatically?"
}
}
}'The same request costs $0.0000026 on DecisionNode. Compare it with DecisionNode Flash on your own inputs in the playground: the model picker switches between them without changing anything else.
When it has no ready capacity#
Now and then DecisionNode has no ready capacity for a request, for example while more of it is being brought up. Rather than make you wait, the request is answered at once by DecisionNode Flash, with the same questions and the same answer shapes, and billed at DecisionNode Flash's price. The answer says so: model is decisionnode-flash-0.9.1-rc1, and two headers name what happened.
| You send | You get | Billed |
|---|---|---|
Nothing extra ("fallback": true, the default) | A 200 from DecisionNode Flash: model decisionnode-flash-0.9.1-rc1, x-decisionnode-fallback: decisionnode-flash-0.9.1-rc1, x-decisionnode-requested-model: decisionnode-main-0.9.1-rc1 | At DecisionNode Flash's price |
"fallback": false | 529 overloaded_error with Retry-After: the time DecisionNode needs to be ready, or 30 s | Nothing |
Anything, in a workspace whose fallback policy is never | The same 529 as false | Nothing |
- Opt out when the decision needs DecisionNode: a re-check of an unsure DecisionNode Flash answer, or a decision you pinned to
decisionnode-main-0.9.1-rc1. Send"fallback": falseand retry the529afterRetry-After, as for any529. - Or for the whole workspace: an Owner or Admin turns off DecisionNode Flash fallback in the console's settings (the workspace's fallback policy
never); a request's"fallback": truecannot override it. - Log the model that answered. Read
model(orx-decisionnode-fallback) rather than assuming the one you asked for: thresholds tuned on DecisionNode may need their own values on DecisionNode Flash. - DecisionNode Flash has no fallback. A DecisionNode Flash request with no ready capacity is the usual
529.
| Call | With no ready capacity |
|---|---|
POST /v1/decide for decisionnode-main-0.9.1-rc1 or decisionnode | Answered by DecisionNode Flash, unless the request or the workspace opts out |
| Opening a session | Never falls back: 529 with Retry-After; open it again after the wait |
| A batch line | Never falls back: the line waits in the batch's queue, within the batch's 24 hours |
| A fine-tuned model, on either base | Never falls back: 529 with Retry-After. You asked for your own model, so nothing else answers in its place |
| DecisionNode Flash | No fallback: 529 with Retry-After |
curl https://api.decisionnode.com/v1/decide \
-H "Authorization: Bearer $DECISIONNODE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "decisionnode",
"state": "Customer: I was charged twice and nobody has replied for 3 days.",
"questions": {
"route": {
"type": "choice",
"instructions": "Where should this go?",
"criteria": {
"billing": "money",
"bug": "broken",
"account": "login"
}
},
"urgency": {
"type": "score",
"instructions": "How urgent is this?",
"criteria": ["routine", "today", "urgent", "critical"]
},
"refund": {
"type": "truth",
"instructions": "Refund this automatically?"
}
},
"fallback": false
}'