Cursor Router: Intelligent AI Model Routing for Teams, With Savings of Up to 60%

On July 30, 2026, Cursor made Cursor Router generally available. It is an intelligent model router for teams and enterprise customers that, in online A/B tests on millions of requests, delivered frontier-model quality while saving up to 60%.

How Cursor Router works

Cursor Router is a classifier that analyzes each request before a model runs and sends it to the most suitable model. It evaluates the request itself, the context, the difficulty of the task and its domain, and matches those against what Cursor knows about how each model behaves.

The company describes the problem Router solves this way: about 60% of developers in Cursor pick a single model as their everyday "workhorse." As a result, even simple tasks run at frontier prices, and AI spending grows faster than the quality of the output. Router changes that logic: routine work goes to the cheapest models, interface updates go to the model with the best "taste," and complex, long-running tasks go to frontier models with strong reasoning.

• General availability: July 30, 2026, for teams and enterprise customers
• Trained on 600,000+ live requests, optimized for a satisfaction metric (AFC)
• Three modes: Intelligence, Balance, Cost
• Cursor handles hundreds of millions of code requests per week
• Router accounts for cache misses in both training and cost estimation

How much teams save

According to Cursor, in online A/B tests on millions of requests, Cursor Router delivered frontier-level quality while saving 60%. During the early-access period, dozens of companies got the same performance for 30–50% less.

The company gives figures by mode. Auto Intelligence reaches the level of the Fable model on result satisfaction at roughly 60% lower cost for teams, and at the same time raises satisfaction by about 15% relative to Opus 4.8 at nearly the same price. Auto Balance beats Opus 4.8 on satisfaction at roughly 36% lower cost, and against GPT-5.6 Sol it delivers comparable quality at lower spend. The three modes — Intelligence, Balance and Cost — let a team choose its own point on the price–intelligence curve.

Why online tests rather than offline evals

Cursor deliberately measured the router's effectiveness with large online A/B tests instead of offline evals. In the company's argument, offline tests are small, detached from real usage, and fail to capture the extra cost of cache misses when switching models.

"Our goal is to give teams the best performance and experience for every task, without spending more than the work requires," says the

Cursor blog.

Real routing unfolds inside a conversation: an engineer writes code, asks follow-up questions, catches errors and keeps going, often across hundreds of requests in a week. Router has to pick a model and decide when to switch it under exactly those conditions. The classifier is designed for a world where new models ship frequently, so Cursor can update the router quickly as models are released, without retraining the product from scratch.

What this means

Model routers are moving from idea to ready-made enterprise product: teams no longer have to pay frontier prices for routine work, and cost is tied to the real difficulty of the task.

Frequently asked questions

Who can use Cursor Router?

Cursor Router is generally available to Cursor teams and enterprise customers. Previously, it was available to dozens of companies through early access.

How much can you save with Cursor Router?

According to Cursor, in online A/B tests the router delivered frontier-model quality while saving 60%, and in early access, companies got the same performance for 30–50% less.