For AI agents: The documentation index is at 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).

Inference has a usage-based pricing model, so costs scale with your actual usage.

Bring Your Own Models (BYOM)

BYOM model weights are stored in a service-managed, non-accessible Spaces location, and are billed at $5.00 per month. We do not charge you for browsing or managing imported models in Model Catalog. Costs apply only for storing model weights and for using those models with other paid features, such as dedicated inference deployments.

Model Playground

Usage is charged at the same rate as serverless inference.

Serverless Inference

Serverless inference is billed by DigitalOcean for both open-source and commercial models. Prices align with each provider’s published rates for transparency.

The following shows pricing for foundation models available through serverless inference.

Web Search Requests

You are charged $10 per 1000 requests for using web search with serverless inference.

Dedicated Inference

Dedicated Inference is billed per GPU-hour based on the GPU you use.

Batch Inference

Batch inference is charged at up to a 50% discount on OpenAI and Anthropic models.

You are only charged for completed requests. If a batch job fails, is blocked by guardrails, or expires partway through, requests that were not processed are not charged.

Model Evaluations public

Model evaluations for candidate models deployed on Serverless Inference, and for judge models, are charged at the same token rates as serverless inference.

Inference Router public

Inference Router is available in public preview and enabled for all users. You can contact support for questions or assistance.

There is no additional cost to using Inference Router during public preview. Using inference routing forwards requests to foundation models for serverless inference and dedicated inference. You are billed for the models that serve each request.

Knowledge Bases

Knowledge base pricing is shown per million tokens, but billing is calculated per thousand tokens.

You’re billed for both indexing and storage:

• Tokens used for indexing and retrieval query vectorization: We charge for tokens used to generate embeddings during indexing and to vectorize user queries during retrieval. Both use the same embeddings model pricing.

  Indexing pricing is the same for manual and auto-indexing. Indexing charges apply only when changes are detected, such as new, updated, or deleted files or URLs. If auto-indexing is paused or no changes are found, there are no indexing charges.

  Note

  Retrieval requests sent through a MCP server are billed the same as retrieval requests sent directly to the knowledge base retrieve endpoint. This includes the tokens used to vectorize the retrieval query with the selected embeddings model.

  For example, a 10 MB dataset is about 3 million tokens, and a 1 GB dataset is about 250 million tokens.

  Actual costs depend on the embeddings model:

  Model	Price
  all-mini-lm-l6-v2	$0.009 per 1M input tokens
  multi-qa-mpnet-base-dot-v1	$0.009 per 1M input tokens
  gte-large-en-v1.5	$0.09 per 1M input tokens
  Qwen3 Embedding 0.6B	$0.04 per 1,000,000 tokens
  BGE-M3	$0.02 per 1,000,000 tokens
  E5 Large V2	$0.02 per 1,000,000 tokens

  Note

  One token is roughly four characters (approximately 75 words per 100 tokens). Non-Latin scripts, emojis, or binary data may increase token counts.

• Reranking tokens: If reranking is enabled, tokens used to rerank results are billed based on the selected reranking model. For supported reranking models, see available reranking models.

  Model	Price
  BGE Reranker v2 m3	$0.01 per 1M reranking tokens

• Storage: Embeddings are stored in OpenSearch. See OpenSearch pricing.

Chunking has no separate charge. Chunking costs depend on embedding token usage, OpenSearch database, and the selected embeddings model.

Chunking strategy cost depends on how many tokens the strategy embeds and returns:

• Section-based and fixed length chunking are the most cost-efficient because they use simple splitting and have predictable token usage.
• Semantic chunking costs more because it uses the embeddings model to detect semantic boundaries and embed final chunks, often resulting in 1.5 to 3 times more indexing tokens.
• Hierarchical chunking slightly increases indexing cost by creating parent and child embeddings. It can also increase retrieval cost because agents receive both child and parent chunks for each lookup.

Changing your chunking strategy or configuration requires re-indexing the affected data source, which consumes additional tokens. For guidance on chunking configurations and best practices, see our chunking parameters reference and chunking best practices.

If you use RAG Playground, answer generation is billed separately based on the selected serverless inference model. Free tokens for RAG Playground are not separate; they are shared with Model Playground.

Agent Platform

Agent creation is free. We charge for model usage and for additional features like knowledge bases, guardrails, and log stream insights. We display prices per million tokens and bill per thousand tokens for accuracy.

Model usage is billed by DigitalOcean. You are charged for all input and output tokens processed by the agent at the same token rates as serverless inference. Token usage depends on factors such as input length, agent instructions, attached knowledge bases, and configuration settings. To optimize usage, test your agents and adjust their parameters.

Agent Guardrails

Charges apply for all tokens processed through agent guardrails:

Costs are per token. Creating, editing, or duplicating guardrails has no additional cost.

Functions

If you attach DigitalOcean Functions to your agent, you are billed at functions pricing.

Agent Evaluations

Agent evaluations are charged by token usage at the same rates as model usage.

Log Stream Insights

Log Stream Insights uses a third-party model to analyze agent trace data. You are charged per token:

Agent Development Kit public

You are not charged for using the Agent Development Kit during public preview. However, you are billed for other DigitalOcean Inference features you use with your agent deployment:

• We charge for model usage for Agent Development Kit (ADK). If you are using a DigitalOcean-hosted model, you are charged for those model keys.

• For agent evaluations, token usage is charged to the agent model keys. For example, if your agent uses a serverless inference endpoint key, any token usage is charged to that key. If the agent uses a third-party model key, or a key to a model not hosted on DigitalOcean, you are charged by the hosting provider.

• If you enable Log Stream Insights for your agent deployment, you are charged for tokens when new insights are generated.