---
title: DigitalOcean Evaluations Pricing
description: Pricing for dataset evaluations and simulations.
product: Evaluations
url: https://docs.digitalocean.com/products/evaluations/details/pricing/
last_updated: "2026-10-13"
---

> **For AI agents:** The documentation index is at [https://docs.digitalocean.com/llms.txt](https://docs.digitalocean.com/llms.txt). Markdown versions of pages use the same URL with `index.html.md` in place of the HTML page (for example, append `index.html.md` to the directory path instead of opening the HTML document).

# DigitalOcean Evaluations Pricing

DigitalOcean Evaluations measure model, Inference Router, and agent performance using dataset evaluations and simulations.

DigitalOcean Evaluations charges depend on the models and inference resources used during an evaluation.

## Dataset Evaluation Pricing

Dataset evaluations can use a model, Inference Router, or supported agent as the candidate.

Evaluations that use candidate models deployed on Serverless Inference are charged at the same token rates as [serverless inference](https://docs.digitalocean.com/products/inference/details/pricing/index.html.md#serverless-inference-pricing). Judge model usage is charged at the applicable model token rates.

Candidate models deployed on Dedicated Inference don’t incur additional evaluation-specific token charges.

Multi-turn dataset evaluations don’t use a candidate model because the responses being evaluated are already included in the dataset. Judge model usage still applies.

## Simulation Pricing (public)

Simulations don’t have a per-run charge. Instead, you pay for the tokens that the simulation’s models use and for the agent you evaluate:

- **User simulator and judge models**: Input and output tokens are charged at the same token rates as [serverless inference](https://docs.digitalocean.com/products/inference/details/pricing/index.html.md#serverless-inference-pricing), and appear on your invoice as Evaluations token usage. The evaluation configuration displays the input and output token pricing for each available model. Tokens used before a run fails or is cancelled are charged.
- **Scenario generation**: When you [generate a scenario set from a goal](https://docs.digitalocean.com/products/evaluations/how-to/manage-scenario-sets/index.html.md#generate-scenarios-from-a-goal), the generation model’s tokens are charged at the model’s token rates, whether or not you use the scenario set in a run.
- **Agent under test**: The agent’s side of the generated conversations is charged as [Harness Runtime usage](https://docs.digitalocean.com/products/managed-agents/agent-harness-runtime/details/pricing/index.html.md), the same as other managed agent sessions, including its compute, storage, tools, and model inference. This applies to managed agents and to agents that the simulation creates from a model or Inference Router. If the agent uses a third-party model with your own provider key, such as an Anthropic or OpenAI key, DigitalOcean charges only for the agent’s compute, storage, and tools, and the provider bills you directly for inference.

## Agent Droplets Discounts

If you have an [Agent Droplets plan](https://docs.digitalocean.com/products/managed-agents/agent-harness-runtime/details/pricing/index.html.md#agent-droplets), its discount (15% for Pro, 20% for Team) also applies to evaluation charges for DigitalOcean-hosted models:

- **Dataset evaluations**: Candidate model inference and judge model usage.
- **Simulations**: Scenario generation, the user simulator model, the agent under test’s Harness Runtime usage, and judge model usage.

Third-party models, such as Anthropic Claude and OpenAI GPT, are not discounted.

## Storage

Storage for evaluation datasets and results is currently provided at no additional charge. DigitalOcean may introduce or modify storage fees in the future.