DigitalOcean Serverless Inference

What it does

DigitalOcean Serverless Inference lets developers send prompts, images, audio, and text-to-speech requests to hosted foundation models through API endpoints without managing GPU infrastructure. Requests go through chat completions, responses, image, audio, and video generation interfaces, billed by usage at each model's published rate. A Model Catalog lets you compare model capabilities and pricing, and Inference Router can route requests to a suitable model.

Teams typically create model access keys, retrieve available models, and call the endpoints from their own code or agent applications. Documented features include prompt caching, reasoning options, batch inference at up to a 50% discount on OpenAI and Anthropic models, and knowledge base retrieval. Dedicated Inference, a separate per-GPU-hour offering, is available for reserved hardware.

This suits developers and engineering teams who want usage-based access to a range of open-source and commercial models. Serverless inference is prepaid only: you need promotional credits or a positive balance, and access is suspended when funds run out. Some tools, such as web search and web fetch, carry extra per-request charges.

What it can do

• Chat Completions and Responses API endpoints for text generation with reasoning options
• Prompt caching across Chat Completions, Messages, and Responses APIs
• Image generation and editing from text prompts
• Multimodal processing across images, audio, video, and text
• Text-to-speech and audio generation through fal models
• Model Playground and RAG Playground for testing

Works with

• Model Context Protocol (MCP) servers, including DigitalOcean MCP servers
• Knowledge bases with embeddings stored in OpenSearch

Before you choose

• Serverless inference is prepaid only; access is suspended when credits and balances run out.
• Prices are per million tokens for text models and per request or per compute second for image, audio, and video models; rates change over time, so check the pricing page.
• Inference Router is in public preview with no additional cost, but you pay for the models that serve each request.

Listed under

• Coding
• Sandboxes and infrastructure

Get started

Create an account, fund the Inference and Agents balance or prepayment balance, then create a model access key.

1. Create a DigitalOcean account and fund the Inference and Agents balance or account prepayment balance, or use promotional credits.
2. Create a model access key scoped to the models or routers you need.
3. Retrieve available models and send requests to the chat completions or responses endpoint from your code.

Prepaid only. Promotional credits or a positive balance are required. Batch inference can be up to 50% off OpenAI and Anthropic models. Dedicated Inference is billed per GPU-hour separately.

Source: docs.digitalocean.com/products/inference/details/pricing, checked 29 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• Use Serverless Inference | DigitalOcean Documentation ↗docs.digitalocean.com
• Inference Pricing | DigitalOcean Documentation ↗docs.digitalocean.com
• AI-Native Cloud | DigitalOcean ↗www.digitalocean.com

Company website

docs.digitalocean.com: product, pricing and policy pages

Web search

Confirms the official site

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

manual_review

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.