> ## Documentation Index
> Fetch the complete documentation index at: https://narev.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Create a benchmark

> Walk through the four steps to create a benchmark in Narev Cloud, attach a data source, add variants, and start comparing LLM performance.

You need a Narev account to create a benchmark. You can create one by registering [here](https://narev.ai/login).

With an account, you can create a benchmark in four steps.

<Steps>
  <Step title="Create the benchmark">
    Benchmarks and all of their management options are in **Narev Cloud -> Benchmarks**. After you create a benchmark, the to-do list appears.

    <Frame>
      <img src="https://mintcdn.com/narev/Y1illty3WLN-yruS/images/benchmark/benchmark-setup-annotated.png?fit=max&auto=format&n=Y1illty3WLN-yruS&q=85&s=c321337e4e6938db57dc9a360b84bb57" alt="Narev Cloud benchmark setup checklist view" width="1496" height="1352" data-path="images/benchmark/benchmark-setup-annotated.png" />
    </Frame>
  </Step>

  <Step title="Define the data source">
    There are multiple ways to add a data source to a benchmark, from one-off manual entry to programmatic ingestion through an API endpoint.

    Each data source has a separate page:

    * [Manual entry](/platform/benchmark/data-source/manual-entry)
    * [File upload](/platform/benchmark/data-source/file-upload)
    * [Live test](/platform/benchmark/data-source/live-test)
    * [Tracing platform](/platform/benchmark/data-source/tracing-platform)
    * [Clawhub](/platform/benchmark/data-source/clawhub)
    * [Using API](/platform/benchmark/using-api)

    <Frame>
      <img src="https://mintcdn.com/narev/Y1illty3WLN-yruS/images/benchmark/benchmark-datasource-annotated.png?fit=max&auto=format&n=Y1illty3WLN-yruS&q=85&s=fd59d5b05a7852b36f385c23612fd4ff" alt="Narev Cloud benchmark data source selection" width="1928" height="818" data-path="images/benchmark/benchmark-datasource-annotated.png" />
    </Frame>
  </Step>

  <Step title="Configure variants">
    To run a benchmark, you need to configure at least one [variant](/platform/concepts/variants). You can select variants with the checkbox marked as `1`.

    `2` allows you to create a variant from scratch and `3` gives you recommendation based on the currently selected variants.

    <Frame>
      <img src="https://mintcdn.com/narev/Y1illty3WLN-yruS/images/benchmark/benchmark-variants-annotated.png?fit=max&auto=format&n=Y1illty3WLN-yruS&q=85&s=d0a0c58923daa9e439cd12319dbaae1f" alt="Narev Cloud benchmark variant selection view" width="1930" height="1430" data-path="images/benchmark/benchmark-variants-annotated.png" />
    </Frame>
  </Step>

  <Step title="Configure quality evaluation">
    Each benchmark requires a way to evaluate quality.

    There are multiple ways to evaluate quality. For concepts, see [Quality evaluations](/platform/concepts/quality-evaluations).

    <Frame>
      <img src="https://mintcdn.com/narev/Y1illty3WLN-yruS/images/benchmark/benchmark-quality-metrics-annotated.png?fit=max&auto=format&n=Y1illty3WLN-yruS&q=85&s=081582c5d0c3ad4599ca85dc9f887d63" alt="Narev Cloud quality metrics setup view" width="1876" height="1302" data-path="images/benchmark/benchmark-quality-metrics-annotated.png" />
    </Frame>
  </Step>
</Steps>

With this setup, you are ready to run your first benchmark.