Serverless Inference 
55+ AI models. One API. Zero infrastructure to manage.

Run frontier and open-source AI models through a single endpoint with built-in routing, multimodal generation, and pay-per-token pricing.

No GPUs to provision · No deployments to configure · No idle infrastructure costs

Unified Platform
More than an inference API.

Serverless Inference lets you use frontier and open-source models through a single platform without managing separate vendors, APIs, or infrastructure. It combines intelligent routing, multimodal generation, observability, and built-in cloud infrastructure so you can optimize cost, latency, and reliability without rebuilding your stack. Inference next to your databases, storage, and networking. No egress between layers.

Frontier models

DigitalOcean-hosted

Browse the full model catalog

Built for scale
Built for production AI workloads

• Ranked #1 by Artificial Analysis on output speed for DeepSeek V3.2 and Qwen 3.5 397B
• 3.9x faster than Amazon Bedrock on DeepSeek V3.2
• Zero data retention by default
• Native multimodal support for image, video, audio, and vision-language workloads
• Automatically route requests across models to optimize cost, latency, and reliability with Inference Router, now in Public Preview
  

Up and running in minutes
Three steps and you're making API calls.

1. Generate a Model Access Key
Create and scope a key from your DigitalOcean control panel.

2. Pick your model
Choose from 55+ frontier and open-source models for your use case.

3. Start making API calls
Works with the tools you already use, including OpenAI SDK, cURL, and the DigitalOcean API.

Read the quickstart

Full docs  ·  Pricing  ·  Prompt caching  ·   Inference Router

Follow us

Copyright DigitalOcean. All rights reserved.
105 Edgeview Dr., Suite 425, Broomfield, CO 80021