Run a Judge Over a Whole File

Import your rows, decide what to ask about each one, and get back a structured answer per row — with the probability behind it, not just a label. Results stream in as they land and export as CSV or JSONL.

Up to 1000 rows per batch·Pause and resume any time·Failed rows are never charged

Import your data

CSV, TSV, TXT or JSONL. Up to 2MB and 1000 rows per batch.

Drag a file here, or choose one

One row in, one decision out.

Choose your judge

The questions every row is asked. Save a judge to reuse it on the next file.

Start from a scenario

Saved judges

Model

Question id

Type

Instructions

Answers with one of your labels, plus the probability of each.

Labels

Question id

Type

Instructions

Answers with a position on your ordered scale, plus a distribution.

Tiers, lowest first

Question id

Type

Instructions

Answers with a probability between 0 and 1 that the statement holds.

Question id

Type

Instructions

Answers with a probability between 0 and 1 that the statement holds.

Save this judge as

Saved privately in this browser. Saving is free and runs no model.

Run and export

Results appear row by row. Pause any time; a paused batch resumes where it stopped.

1 credit per row. If credits run out, actual input tokens are used instead. Output tokens are free, and a row that fails is not charged.

Pricing and Limits

1 credit per row. If credits run out, actual input tokens are used instead. Output tokens are free. Oversized input returns an error without a charge and is never silently truncated; an unavailable model stops the batch before anything is reserved.