Get started
All guides
THE ALLEVALS FIELD GUIDE

Getting started with All Evals

Choose a benchmark, inspect its scoring rules, and import it into your evaluation platform.

Published

Preserved original guide. References to Langfuse or Braintrust describe the legacy import product, not a promise that every catalog entry is runnable in the new platform. See the current endpoint evaluation docs and each benchmark’s availability.

All Evals packages public AI benchmarks for use in Langfuse and Braintrust. Each benchmark page explains what the dataset tests, which rows are runnable, and how answers are scored.

Choose a benchmark

Browse the benchmark library and open a dataset that matches the capability you want to measure. Check its scope and score caveat before comparing results with published leaderboards: some packages use adapted prompts or judges.

Import and run

From the benchmark page, choose the import flow for Langfuse or Braintrust. The page explains what arrives in your project and any setup needed to run an experiment. After importing, test a small set of items first so you can review model outputs and grading behavior before a full run.

Read the score in context

A score only describes the selected rows, prompt, model setup, and grader. Keep those details alongside your results when you compare runs.

Browse benchmarks Try the endpoint API