DigitalOcean Inference Router Enhancements: Performance Tuned Routing, No Rules to Write

Every few weeks, a new model steps up and claims best-in-class at something. The launch chart makes the case, the link makes the rounds, and your team is back to the question nobody has time for: should we switch?

The answer takes a week of evals, side-by-side tests and debate, and by the time your team lands on one, the next contender is already on the scoreboard. It’s a race you can’t win manually. And the launch charts only tell part of the story. They show where a model is strongest, not where it isn’t.

Today, we’re enhancing the DigitalOcean Inference Router so it makes that call for you. No presets to select, no custom routers to build. Send your prompt, set the model to inference-router, and choose a preference if you want one. The router sends each request to the model that performs best for that job, validated by public benchmarks.

The right model for the job, automatically

Under the hood, inference-router draws on the entire DigitalOcean Model Catalog: frontier models from Anthropic and OpenAI, open source models like Kimi, GLM, and DeepSeek, all one request away. Every new model gets the same test. The ones that measure up are added to the router.

And because every request goes to the model that fits the job, you stop paying frontier prices for work a leaner model handles just as well. In our internal testing across software engineering, reasoning and agentic work, Inference Router had higher accuracy than Claude Opus 5.5 and GPT-6 Astra, at about 22% lower cost.

With Inference Router, you stop:

• Re-running evals every time a model launches
• Rewriting routing rules when the lineup changes
• Redeploying to swap a model
• Holding the “should we switch?” meeting

The Inference Router Getting Started page on DigitalOcean cloud console shows the benchmarks behind each use case, how the router scores against the best single model, and which models it’s choosing between. It updates whenever we add a model or rerun an evaluation.

Here is how some of the models behind the router score on some of the benchmarks we use:

Scores are from DigitalOcean’s internal evaluations as of October 12th, 2026. Results vary with configuration and may differ from public leaderboards.

How it works

We measure models across a wide variety of tasks and evaluate how well they did, along with the cost and latency. That evidence decides who wins each task.

When a request arrives, the routing model reads it, matches it to the right task, and weighs your preference to pick the best model. The router is intrinsically cache-aware and knows to hold onto a model to re-use its cache unless a switch pays for itself.

As all the heavy lifting happens before your request arrives, you don’t have to manage the configuration manually. When a newly released model does better on a specific task, you benefit automatically.

The router is initially trained on the following task categories:

• Software engineering
• Agentic workflows
• General intelligence and reasoning

While the Inference Router measures well on public benchmarks & our internal evals, that doesn’t tell you how well it does on yours. So while we are starting with a generally optimized router, soon you’ll be able to bring in your own rubric. Bring your evaluation set, and your definition of success decides which model wins.

Set a preference. We handle the rest.

Inference Router is managed end to end by DigitalOcean, so most requests won’t need any input from you.

When you want to influence the choice, use budget to nudge the router to prefer lower cost models or quality for when success is critical. However, we expect routing_preference: balanced (the default) to work for most customers.

Preferences are the first dial. Soon, your own evaluation set will steer the router too, and when a new model ships, we’ll re-run them so you stay on the best fit. Want early access? Request it from the Getting Started page through the DigitalOcean Console.

Get started

Inference Router is available today in Public Preview. There is no additional cost to use it during Public Preview; you’re billed for the models that serve each request.

1. Navigate to Inference Router in the DigitalOcean Control Panel.
2. Generate a model access key if one is not already configured.
3. Send API calls to https://inference.do-ai.run/v1 with inference-router as the model.

curl https://inference.do-ai.run/v1/chat/completions \
  -H "Authorization: Bearer $MODEL_ACCESS_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "inference-router",
    "preference": "balanced",
    "messages": [{"role": "user", "content": "Fix the failing test in ..."}]
  }'