---
title: "Browser Benchmark: Browserbase, Steel, Kernel and More"
description: "Stealth Bench V2: 100 bot-protected sites, 2 visits each. Browser Use got in on 90.5%, Anchor 82%, Kernel 80.5%, Browserbase 78.5%. Raw data on GitHub."
canonical: "https://browser-use.com/benchmarks/browsers"
last_updated: "2026-11-12"
---

# Browser Benchmark: Browserbase, Steel, Kernel and More

## The browser that gets through.

100 bot-protected sites. 2 visits each. **90.5%** of our visits got in. Next best: 82%.

Visits that got in _Sep 25, 2026_

1.  Browser Use
    
    **90.5%**
    
2.  Anchor
    
    **82%**
    
3.  Kernel
    
    **80.5%**
    
4.  Browserbase
    
    **78.5%**
    
5.  Browserless
    
    **77.5%**
    
6.  Steel
    
    **74.5%**
    
7.  Hyperbrowser
    
    **73.5%**
    
8.  Chrome + US proxy
    
    **14%**
    
9.  Headless Chrome
    
    **3%**
    

[Models](https://browser-use.com/benchmarks/agents)[Browsers](https://browser-use.com/benchmarks/browsers)[Shopping](https://browser-use.com/benchmarks/consumer)

## Why we get through.

### Our own Chromium

**84%** Solver off. Still beats every rival.

### Home IPs

**$5/GB** Residential, pay as you go.

### CAPTCHA solver

**+6.5** points: 84% to 90.5%.

## Site by site.

Got into **94 of 100** sites. Next best: 89.

One square per site, same order in every grid. In both visitsOneBlockedWe got in, they didn't

Browser Use **90.5%**

Anchor **82%**

Kernel **80.5%**

Browserbase **78.5%**

Browserless **77.5%**

Steel **74.5%**

Hyperbrowser **73.5%**

Chrome + US proxy **14%**

Headless Chrome **3%**

## How we test.

1.  100 bot-protected sites
    
2.  2 visits each, 60 s
    
3.  Blind judge
    

Sites picked where we got through before. Rivals on standard plans. Sep 25, 2026. [Runner and raw data →](https://github.com/browser-use/benchmark/tree/74b427184d84b0df391a9c337def5aadf115a893/stealth_bench)

## FAQ

Which cloud browser gets blocked least?

Browser Use: in on 90.5% of visits to 100 bot-protected sites. Next best, Anchor: 82%.

How is it measured?

Each browser visits each site 2 times, 60 seconds per visit. A judge model sees the page, not the browser's name, and decides if the real site loaded.

Are the sites a fair sample?

No. We picked sites where Browser Use got through before and Chrome with a US proxy did not. They are hard on purpose.

Was it rerun?

Yes, independently, on Sep 25. Browser Use 90%, best other 85%.

Can I check the data?

Yes. The runner and every visit's result are [on GitHub](https://github.com/browser-use/benchmark/tree/74b427184d84b0df391a9c337def5aadf115a893/stealth_bench).

## Start with $1 free.

One per person. Never expires.

[Get API key](https://cloud.browser-use.com/settings?tab=api-keys&new=1)

---

BROWSER USE

- Browser Use Agents: give Browser Use a task and receive completed work. API V4 is current for new integrations.
- Browser Infrastructure: connect your agent or automation to managed browsers through SDK, REST, or CDP. Starts at $0.06/browser-hour.
- Developer tools: Open Source, Browser Harness, SDK, and MCP support the two products above.

[Developer Index](https://browser-use.com/index.md)
[Product Map](https://browser-use.com/llms.txt)
[Full Product Context](https://browser-use.com/llms-full.txt)
[Pricing](https://browser-use.com/pricing.md)
[Cloud Docs](https://docs.browser-use.com/cloud/quickstart)
[Open Source Docs](https://docs.browser-use.com/open-source/introduction)
