All articles

BU Ultrafast: a white craft racing over red canyons, median task 22s, $0.006

BU Ultrafast is our fastest browser agent. Until Oct 30 it was behind a waitlist. Now it is on every Browser Use Cloud plan in the US, free plan included.

22 seconds, under a cent

The median task takes 22 seconds and costs $0.0064. Our default agent on GPT-6 Luna takes 79 seconds and $0.0098. That is more than 3x faster, and cheaper per task.

Source: medians of completed runs, Oct 16-29, 2026, at most 1,000 per project. Ultrafast: 12,238 runs, 134 projects. Default: 146,120 runs, 9,473 projects. In the 8 projects with 20+ runs on both, Ultrafast was faster and cheaper in all 8. Cost is the model and search charge per run; browser and proxy are billed separately.

Try it

Set model to bu-ultrafast. Nothing else changes.

curl https://api.browser-use.com/api/v4/runs \
  -H "X-Browser-Use-API-Key: $BROWSER_USE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"task": "Find the cheapest flight from SFO to JFK next Friday", "model": "bu-ultrafast"}'

Need more reasoning per step? Use bu-fast. It is the same agent.

Pricing

bu-ultrafast runs GPT-6 Luna on OpenAI's priority tier, so tokens bill at priority rates, 2x standard. bu-fast bills at standard rates, and so does bu-ultrafast on HIPAA projects. Both work on the free plan.

If you joined the waitlist, there is nothing to do. You already have it.

Read more: BU Ultrafast, the docs, or start in Browser Use Cloud.

Published October 31, 2026

Keep reading

All articles

Tutorials

/Oct 31, 2026

How to Track Competitor Prices with Python

Build a Python price tracker with Browser Use and SQLAlchemy. Scrape three stores, save prices in SQLite, detect changes, and schedule daily checks.

Nik Tomazic

Comparisons

/Oct 27, 2026

Browser Use vs Browserless: Pricing, Stealth & Features

Compare Browser Use and Browserless on pricing, stealth, and performance. Find the right browser infrastructure for your automation workflows.

Nik Tomazic

Engineering

/Oct 27, 2026

Odysseys2: a long-horizon web agent benchmark

100 long tasks on the live web, from Carnegie Mellon University and Browser Use. The best agent fully completes 47 of 100.

Magnus Müller & Lawrence Jang