|

Cursor Releases Cursor Router: A Request-Level Classifier Delivering Frontier Coding Quality at 30–50% Lower Cost

ByRicardo

Cursor has made Cursor Router usually accessible for Teams and Enterprise plans. The system is a classifier that inspects every request earlier than a mannequin runs, then dispatches it to the mannequin finest suited to that particular activity. The cursor workforce reviews frontier-quality efficiency at 60% financial savings in on-line A/B exams, and 30–50% financial savings for early-access enterprise accounts.

The downside it targets is a spend sample fairly than a functionality hole. Cursor states that roughly 60% of its builders decide a single mannequin as a day by day driver. Routine work subsequently will get accomplished at frontier costs, and AI spend grows sooner than output high quality. Routing is Cursor’s reply to that mismatch.

What the classifier really reads

Cursor Router just isn’t a fallback chain or a retry mechanism. It is a classifier skilled on 600k+ stay requests, evaluated in a web based A/B take a look at throughout hundreds of thousands of stay requests, and optimized for consumer satisfaction (AFC) as its reward sign.

For every request, the router analyzes 4 inputs: question, context, activity complexity, and area. It combines these with realized information of every mannequin’s habits. Cursor publishes three routing guidelines that comply with from that classification:

  • Simple work goes to essentially the most price-efficient fashions.
  • UI updates go to the mannequin with the perfect style.
  • Complex, long-horizon issues go to frontier reasoning fashions.

That third rule carries the load of the price argument. The financial savings don’t come from downgrading onerous issues. They come from eradicating routine work from frontier pricing whereas the troublesome tier stays intact.

One implementation element deserves consideration from anybody who has constructed a router. Cursor Router is cache-aware in each coaching and analysis. It is skilled on a dataset the place routing produces cache misses, and the reported price financial savings embody the price of these cache misses. Switching fashions mid-conversation invalidates immediate cache, and that price is actual. Routers that ignore it overstate their financial savings.

The classifier was additionally designed for mannequin churn. Cursor states it might replace the router as newer fashions ship, which issues in a market the place the frontier strikes month-to-month.


Interactive Explainer

Inside Cursor Router: classification, tiers and revealed outcomes

Every determine under is taken from Cursor’s launch publish and changelog (July 22, 2026). Values Cursor didn’t publish are marked as such.




Pick a activity — see how the classifier reads it

Step 1 · Signals learn earlier than any mannequin runs

▼

Step 2 · Routing tier chosen

Cursor publishes three routing guidelines and doesn’t identify a particular mannequin per activity sort. The tier proven is Cursor’s personal revealed class; the pattern duties are illustrative purposes of these three guidelines. Cursor states the classifier reads question, context, activity complexity and area, mixed with what it is aware of about every mannequin’s habits.

Auto mode — three factors on the price–intelligence Pareto frontier


Cursor’s description

Published price per commit (early entry measurement)

Linear projection — drag to set month-to-month commit quantity

Volume: 1,000 commits / month
Model / mode $ / commit Monthly vs Opus 4.8
Cost-per-commit figures are Cursor’s revealed early-access observations. The projection is straight multiplication of these unit figures — it’s arithmetic, not a Cursor forecast, and assumes a continuing activity combine. Cursor reviews GPT-5.6 Sol matched the price of Intelligence mode however with decrease consumer satisfaction; it didn’t publish an actual determine for it.

Element What Cursor reviews
Training information 600k+ stay requests
Reward sign User satisfaction (AFC)
Evaluation Online A/B take a look at throughout hundreds of thousands of stay requests, not offline evals
Quality metric 1 User satisfaction — agent success categorized from consumer responses. Moving to the following function is a powerful constructive; correcting the agent is a powerful damaging.
Quality metric 2 Keep fee — how a lot agent-generated code stays within the codebase over time
Metric monitor document Both metrics used to judge each mannequin launch and harness change for the previous 9 months
Cache dealing with Cache-aware in coaching and analysis. Trained on a dataset the place routing causes cache misses; reported financial savings embody cache-miss price.
Platform scale Cursor routes a whole lot of hundreds of thousands of coding requests per week throughout fashions and suppliers
Problem being solved ~60% of Cursor builders decide one mannequin as a day by day driver, placing routine work on frontier pricing
Cursor’s acknowledged purpose for selecting on-line A/B exams: offline evals are restricted by small dimension, distance from actual utilization, and the issue of decreasing success to a rubric — they usually omit the additional cache-miss price incurred when switching fashions mid-conversation.

Source: cursor.com/blog/router · verified Jul 22, 2026
Built by Marktechpost

Similar Posts