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

# OpenGround

> Compare multiple LLMs side-by-side on cost, response time, and output quality using OpenLIT's built-in playground

OpenGround is OpenLIT's LLM playground for side-by-side comparison. You configure two or more LLM providers, enter a prompt, and run them simultaneously. OpenGround displays each response alongside its cost, duration, and token counts so you can make data-driven decisions about which model to use.

## Features

* **Side-by-side comparison** — run multiple LLMs against the same prompt simultaneously and view their responses in parallel
* **Performance metrics** — compare response time and token usage (prompt tokens, completion tokens) for each model
* **Cost analysis** — see the estimated cost of each request to help you balance budget and performance
* **Response quality** — read and compare the full text of each model's response for the same input
* **Experiment history** — all experiments are saved automatically so you can revisit and share previous comparisons

## Access OpenGround

In the OpenLIT UI, click **OpenGround** in the left navigation sidebar. The landing page shows all previously created experiments. Click any experiment to open its results, or click **Create new** to start a fresh comparison.

## Run an experiment

<Steps>
  <Step title="List existing experiments">
    When you open OpenGround, you see an overview of all previously created experiments. Each row shows the experiment name, the models compared, and when it was last run.

    Browse existing experiments to find prior comparisons before creating a new one — you may have already run the comparison you need.
  </Step>

  <Step title="Create a new experiment">
    Click **Create new** to open the experiment editor.

    Configure your first LLM:

    1. Choose a **provider** (e.g., OpenAI, Anthropic, Mistral)
    2. Select a **model** from that provider
    3. Set model **parameters** such as temperature and max tokens
    4. Enter the provider **API key** to authorize requests

    Then configure your second LLM by repeating the same steps for a different provider or model.

    <Tip>
      You can store LLM API keys in [OpenLIT Vault](/openlit/secrets/vault) and retrieve them centrally instead of entering them each time you run an experiment.
    </Tip>
  </Step>

  <Step title="Enter your prompt and compare">
    Once both models are configured, type your prompt into the prompt editor and click **Compare Response**.

    OpenGround sends the prompt to both models in parallel and displays the results side-by-side as they arrive.
  </Step>

  <Step title="Review the results">
    For each model, OpenGround shows:

    | Metric              | Description                               |
    | ------------------- | ----------------------------------------- |
    | **Response**        | The full text output from the model       |
    | **Duration**        | Total response time in milliseconds       |
    | **Cost**            | Estimated USD cost based on token pricing |
    | **Prompt tokens**   | Number of tokens consumed by the input    |
    | **Response tokens** | Number of tokens in the model's output    |

    Use these metrics together to evaluate the trade-off between response quality, latency, and cost for your specific use case.
  </Step>
</Steps>

## Comparison metrics

OpenGround surfaces three categories of metrics to help you choose between models:

**Cost** — estimated cost per request based on each provider's per-token pricing. Use this to identify which model fits your budget, especially when running high-volume workloads.

**Duration** — end-to-end response time from request to final token. Use this to evaluate latency-sensitive applications such as chatbots or real-time assistants.

**Response tokens** — the number of output tokens generated. Longer responses consume more tokens and cost more. Compare response length alongside response quality to find models that are concise without sacrificing accuracy.

<Note>
  Cost estimates in OpenGround are calculated from publicly listed pricing at the time the experiment runs. Actual costs may differ based on your provider agreement, caching, or pricing changes.
</Note>

## Experiment history

Every experiment you run is saved automatically. You can return to **OpenGround** at any time to review past results, share them with teammates, or re-run the same prompt against updated model versions.

***

<CardGroup cols={2}>
  <Card title="Evaluations" href="/openlit/evaluations/overview" icon="check-circle">
    Score LLM outputs for hallucination, bias, and toxicity automatically
  </Card>

  <Card title="Manage LLM secrets" href="/openlit/secrets/vault" icon="vault">
    Centrally store LLM API keys for use across experiments without re-entering them
  </Card>
</CardGroup>