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 $0.02/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) --- # Consumer browser agent benchmark: 502 real web tasks ## Consumer benchmark 502 real web tasks. 363 shopping carts. 139 public information tasks. [Method & data](https://browser-use.com/benchmarks/consumer#method) All tasksAll tasksShopping cartsShopping cartsPublic infoPublic information 502 tasks in this view ## Historical results Task completion ### Browser Use 91.4% 459/502 tasks Result details Supplied reviewed result table. Different runner and browser settings remain. 43 failures, 0 policy holds, 0 unknown, 0 unstarted. Source: Cloud workflow 39369520482; paired OpenAI workflow 39352995121 ### OpenAI browser runner 72.3% 363/502 tasks Result details Supplied reviewed result table. Different runner and browser settings remain. 139 failures, 0 policy holds, 0 unknown, 0 unstarted. Source: Cloud workflow 39369520482; paired OpenAI workflow 39352995121 ### Meta Muse 61% 306/502 tasks Result details Muse reports evidence review. Root independently reviewed 37 original tasks, with 24 endpoints and 13 misses. 165 failures, 29 policy holds, 2 unknown, 0 unstarted. Source: Original Muse consumer run, October 21-22, 2026 Historical V8 outcomes on the retained V9 task list. These are not fresh V9 runs or website block rates. ## Runs without a final score ### Instinct 12 endpoints reported, 481 tasks unstarted. Partial Result details Actual result rows and archive structure verified. Original screenshot and action review remains separate. 4 failures, 5 policy holds, 0 unknown, 481 unstarted. Progress across the planned set: 12/502 (2.4%). This is not a final score. Instinct's stricter self-check counts 4 successes. This is a separate provider assessment. Source: Instinct supplied v4 archive, October 23, 2026 ### Grok Bot No verified matching result yet. Pending Result details No matching complete result is available. ### Dot No verified matching result yet. Pending Result details No matching complete result is available. Updated Oct 23, 2026. Expand an agent for evidence and run limits. [Download results](https://browser-use.com/benchmarks/consumer-v9/results.json) [Agents](https://browser-use.com/benchmarks/agents)[Browsers](https://browser-use.com/benchmarks/browsers)[Consumer tasks](https://browser-use.com/benchmarks/consumer) ## Method & data A successful task needs the requested endpoint. An add button alone does not prove a cart. [Benchmark source](https://github.com/browser-use/benchmark/tree/main/consumer_bench) Scoring and run conditions We recalculated historical V8 results on the retained V9 task IDs. Fleet Farm has a corrected URL and task text. These are historical projections, not new V9 executions. Models, dates, browsers, solver rules, and evidence review differ. We selected Browser Use's best retained run. Policy holds and browser permission failures do not establish website blocks. Bars show successful endpoint reports divided by all 502 planned tasks in this view. Missing and unstarted tasks never become website failures. 11 documented exclusions shopee.vnduplicate shopee site shopee.co.idduplicate shopee site shopee.co.thduplicate shopee site shopee.com.myduplicate shopee site very.co.uklocal access denied darty.comlocal access denied webuy.comlocal access denied snapdeal.comlocal access denied goldapple.rulocal access denied livemaster.rulocal access denied buyee.jptemporary maintenance [Task manifest](https://browser-use.com/benchmarks/consumer-v9/dataset-v9.json) [Full method](https://browser-use.com/benchmarks/consumer-v9/methodology.txt) Original run records stay separate from task data.[Synthetic design preview](https://browser-use.com/lab/consumer-benchmark)