---
title: "Create Prediction"
method: POST
path: "/api/v2/arbiters/predictions"
---

# Create Prediction

`POST /api/v2/arbiters/predictions`

Run a single arbiter against a single example and return a flat result.

Unlike v1's ``POST /api/v1/arbiters/predictions`` (which takes a list of
arbiters and wraps the response in a ``predictions`` array), v2 takes
exactly one arbiter and returns ``{example_id, prediction_id, output,
reasoning, messages}``.

## Headers

- `authorization` string, nullable

## Cookies

- `modaic_session` string, nullable

## Request body

- SrcApiV2ArbitersSchemasPredictExampleRequest — v2 single-arbiter, single-example prediction request. Differs from v1 ``PredictExampleRequest`` (which takes a list of arbiters) by accepting exactly one arbiter and putting ground-truth on the example rather than on the arbiter entry.
  - `input` object, required
  - `alt_id` string, nullable
  - `arbiter_repo` string, required
  - `arbiter_revision` string
  - `ground_truth` object, nullable
  - `ground_reasoning` string
  - `compute_confidence` boolean

## Response `200`

Successful Response

- PredictExampleResponse
  - `example_id` string, required
  - `prediction_id` string, required
  - `output` object, nullable
  - `reasoning` string, nullable
  - `messages` object[], nullable

## Other responses

- `422` — Validation Error

---

[API](https://skmtc.net/modaic-ai/apis/modaic-api.md) · [All operations](https://skmtc.net/modaic-ai/apis/modaic-api/llms.txt) · [OpenAPI document](https://skmtc-service-staging.skmtc.workers.dev/v1/apis/modaic-ai/modaic-api/revisions/43a90513ac59/schema)
