https://api.decisionnode.com /v1/modelscurl https://api.decisionnode.com/v1/models \
-H "Authorization: Bearer $DECISIONNODE_API_KEY"Headers#
AuthorizationstringrequiredBearer dn_live_.... The request has no body and no parameters; the list depends only on your key's workspace. It answers a workspace that is out of credit or suspended too, and it is not rate limited.
Call it once at startup to check a key and to log which pinned version decisionnode answers from. It sends no tokens, so it costs nothing. The answer carries Cache-Control: private, max-age=60.
Response body#
{
"models": [
{
"name": "decisionnode",
"description": "Our most accurate decision model",
"release_date": "2026-10-01",
"display_name": "DecisionNode",
"short_name": "DecisionNode",
"version": "decisionnode-main-0.9.1-rc1",
"current_version": "decisionnode-main-0.9.1-rc1",
"alias": "decisionnode",
"price_per_million_input": 0.042,
"price_per_million_output": 0,
"price_per_million_image": 0.09,
"price_per_million_video": 0.09,
"batch_price_factor": 0.7,
"context_tokens": 65536,
"max_images": 16,
"max_choices": 255,
"max_levels": 10,
"number_one_question_values": 256,
"number_max_grid_values": 4096,
"number_default_range": [0, 255],
"number_default_step": 1,
"upload_max_bytes": 10485760,
"upload_ttl_seconds": 3600,
"upload_media_types": ["image/jpeg", "image/png", "image/webp"],
"videos": { "max_videos": 4, "max_frames_request": 128 },
"batch": {
"requests_per_batch": 10000,
"batch_bytes": 268435456,
"add_bytes": 10485760,
"active_batches_per_key": 10,
"queued_requests_per_key": 100000,
"time_to_completion_hours": 24,
"result_retention_days": 7,
"max_custom_id_chars": 512
}
},
{
"name": "decisionnode-flash",
"description": "The fastest and cheapest tier",
"release_date": "2026-10-01",
"display_name": "DecisionNode Flash",
"short_name": "DecisionNode Flash",
"version": "decisionnode-flash-0.9.1-rc1",
"current_version": "decisionnode-flash-0.9.1-rc1",
"alias": "decisionnode-flash",
"price_per_million_input": 0.035,
"price_per_million_output": 0,
"price_per_million_image": 0.08,
"price_per_million_video": 0.08,
"batch_price_factor": 0.5,
"context_tokens": 65536,
"max_images": 16,
"max_choices": 255,
"max_levels": 10,
"number_one_question_values": 256,
"number_max_grid_values": 4096,
"number_default_range": [0, 255],
"number_default_step": 1,
"upload_max_bytes": 10485760,
"upload_ttl_seconds": 3600,
"upload_media_types": ["image/jpeg", "image/png", "image/webp"],
"videos": { "max_videos": 4, "max_frames_request": 128 },
"batch": {
"requests_per_batch": 10000,
"batch_bytes": 268435456,
"add_bytes": 10485760,
"active_batches_per_key": 10,
"queued_requests_per_key": 100000,
"time_to_completion_hours": 24,
"result_retention_days": 7,
"max_custom_id_chars": 512
}
}
]
}modelsarray- One entry per name your key may send.
namestring- What you send as
model: a family id, a pinned id, or one of your workspace's fine-tuned models. descriptionstring- What the model is for.
display_namestring- The model's name for people, such as
DecisionNode. Show this in your own UI; sendname. short_namestring- A short name for people, such as
DecisionNode Flash. release_datestring- When this name started answering, as
YYYY-MM-DD; a pinned entry carries its family entry's. versionstring- The pinned version the name answers from today: on a family's entry, its current version (else the newest our servers report, else the family id itself); on a pinned id's own entry,
nameitself. Send it asmodelto pin. current_versionstring or null- Family entries only: the pinned version as
versionnames it, ornullwhile no server reports one. aliasstring- The family id, such as
decisionnode: on a family's entry its own id, on a pinned entry the family the version belongs to. Fine-tuned entries carry none. price_per_million_inputnumber- The price per million input tokens, in USD.
price_per_million_outputnumber- Always
0: the model writes no text, so output is free. price_per_million_imagenumber or null- The price per million image tokens, in USD;
nullwhen none is set. No fee per image. price_per_million_videonumber or null- The price per million video tokens, in USD;
nullwhen none is set. No fee per frame. batch_price_factornumber- The share of the live prices a batch line pays, from 0 to 1:
0.7on DecisionNode (30% off),0.5on DecisionNode Flash (50% off). scheduledarray- The changes due within 30 days, soonest first, each
{"field", "value", "effective_from"}: a price field and its new value, such as{"field": "price_per_million_input", "value": 0.038, "effective_from": "..."}. A family'sversionis never scheduled: it moves when our servers serve a newer version. Left out when nothing is scheduled. Fine-tuned entries carry none. context_tokensnumber- Maximum tokens per request, 65,536.
max_imagesnumber- Images per request.
max_choicesnumber- Options per choice question.
max_levelsnumber- Levels per score question.
number_one_question_valuesnumber- Grid values a number question answers as one question, 256. Finer or wider grids are composed in the same call.
number_max_grid_valuesnumber- The most values a number question's grid may hold in all, 4,096. A larger grid is a
422. number_default_rangearray- The
minandmaxof a number question that leaves them out,[0, 255]. number_default_stepnumber- The
stepof a number question that leaves it out, 1. upload_max_bytes, upload_ttl_seconds, upload_media_typesnumber, number, array- What an upload of this model takes: its largest picture in bytes, how long it lives in seconds, and its media types.
videosobject- The video limits:
max_videos(4) andmax_frames_request(128), the smaller of ours and the model server's, and whatever else the server publishes about frames. placesobject- Only when the version's servers answer several places:
{"max_places": n}, the most a point or box question may ask for. Absent otherwise. batchobject- The limits of batch jobs on this model; their price is
batch_price_factor.requests_per_batchnumber- Requests one batch may hold, 10,000.
batch_bytesnumber- The largest upload per batch, in bytes (256 MiB).
add_bytesnumber- The largest add call, in bytes (10 MiB).
active_batches_per_keynumber- Batches your key may hold in
draft,queuedorrunning, 10. queued_requests_per_keynumber- Requests your key may hold across its active batches, drafts included, 100,000.
time_to_completion_hoursnumber- Hours after finalize at which unfinished requests expire, unbilled, 24.
result_retention_daysnumber- Days results are kept after a batch ends, 7.
max_custom_id_charsnumber- The longest
custom_id, 512 characters.
fine_tunedboolean- Only on your workspace's fine-tuned models:
true. Our own models' entries do not carry it. basestring- Fine-tuned entries only: the family it was trained from, such as
decisionnode. base_versionstring- Fine-tuned entries only: the pinned base version it was trained on.
familystring- Fine-tuned entries only: the same family id as
base. adapter_versioninteger- Fine-tuned entries only: the version number; on a plain name, the version it answers from.
versionstays the pinned name. statusstring- Fine-tuned entries only:
deployed, orready(trained and passed its gate, not deployed: its name is refused).
curl https://api.decisionnode.com/v1/models \ -H "Authorization: Bearer $DECISIONNODE_API_KEY"Names in the list#
The list holds one entry for every name your key may send. Our own models' entries carry the same fields; a fine-tuned entry differs (below). The sample above shows the first two; the full list today is:
| name | What it is |
|---|---|
decisionnode | DecisionNode, the family. Its version is decisionnode-main-0.9.1-rc1 today and moves with each release |
decisionnode-flash | DecisionNode Flash, the family. Its version is decisionnode-flash-0.9.1-rc1 today and moves with each release |
decisionnode-main-0.9.1-rc1 | DecisionNode, pinned. Its version is its own name and never changes; its alias is decisionnode |
decisionnode-flash-0.9.1-rc1 | DecisionNode Flash, pinned. Its version is its own name and never changes; its alias is decisionnode-flash |
A pinned id is its own entry, whose version equals its name and whose alias names its family. To find what a family id answers from, read its entry's version. The pinned entries are the versions our servers report, at most 16 a family: a version stays listed while a server still reports it (see Versions and retirement). Earlier ids that are still accepted as model are not listed.
The list names the hosted models only. DecisionNode Edge, the open-weight System One models you run on your own hardware, is never listed and is not sent as model: its runtime answers on your machine.
Prices per million input tokens are on Pricing and billing. Code that reads only name, description and release_date keeps working: the other fields are extra.
Read the batch object before you size an upload: code that reads it keeps working when the values change.
A family's version changes when we release a new version; a pinned id never changes. How long a version stays available, and how to move to a new one, is on Versions and retirement.
Fine-tuned models#
For a key of a workspace that has fine-tuned models, the list also holds the plain name of every model with a deployed version, and every deployed or ready version under its pinned name. Only deployed entries answer; a ready one is trained and passed its gate, and sending its name is the 400 Unknown model until you deploy it. Each entry carries the base's limits, as the base's own entry does. Another workspace's keys never see them.
{
"name": "decisionnode-flash@acme/fraud:3",
"description": "Claims fraud, trained on 2,400 decisions",
"release_date": "2026-10-09",
"version": "decisionnode-flash@acme/fraud:3",
"fine_tuned": true,
"base": "decisionnode-flash",
"base_version": "decisionnode-flash-0.9.1-rc1",
"family": "decisionnode-flash",
"adapter_version": 3,
"status": "deployed",
"context_tokens": "65,536",
"max_images": "16",
"max_choices": "255",
"max_levels": "10"
}versionis the pinned name the entry answers from, as on every entry: on the plain namedecisionnode-flash@acme/it is the version deployed most recently.fraud descriptionis the one set in the console, elseFine-tuned fromand the base's name;release_dateis the day the version finished training.- Match on
fine_tuned, not on the shape of the name, to tell your models from ours. - A fine-tuned entry differs: it carries
base_versionand noalias,batch,batch_price_factororscheduled; itsdisplay_nameandshort_nameare your model's name; its prices are its family's current prices plus the serving premium.
Errors#
| Status | Type | When | What to do |
|---|---|---|---|
401 | authentication_error | The key is missing, unknown or revoked | Check the Authorization header |
405 | method_not_allowed_error | Another method than GET; Allow: GET | Send a GET |
500 | api_error | A fault on our side | Retry with backoff |
529 | overloaded_error | No capacity right now | Wait Retry-After, then retry |