WebBench (Halluminate)
Browser agents complete 2,454 tasks on 452 live websites from the global top 1,000 by traffic. Tasks either read information or write data: creating, updating and deleting records, logging in and downloading files. The board compares browser agent products.
It expands WebVoyager from 15 to 452 sites. The README was last updated in May 2025, and the board ranks agent products rather than models. Live websites change, so results drift over time. The README does not say how runs are graded. About 100 GitHub stars.
- Made by
- Halluminate, with Skyvern
- Released
- May 2025
- Kind
- Web browsing
- Level
- Everyday
- Answers
- Agent task
- Tasks
- All public
- Last model added
- January 2025
- Scores refreshed
- November 12th, 2026
Rtrvr AI tops the leaderboard with 81.4%
Best score per entry from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. The best model we track is OpenAI Computer-Using Agent (CUA), at 59.8%.
- 1 Rtrvr AI Human-in-the-loop 81.4%
- 2 OpenAI Operator Human-in-the-loop 76.5%
- 3 Anthropic Sonnet 3.7 CUA Autonomous 66%
- 4 Skyvern 2.0 Autonomous 64.4%
- 5 Skyvern 2.0 on Browserbase Autonomous 60.7%
- 6 OpenAI Computer-Using Agent (CUA) Autonomous 59.8%
- 7 Browser Use Cloud Autonomous 43.9%
- 8 Convergence AI Human-in-the-loop 39.9%
Source: Official WebBench (Halluminate) leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Rtrvr AI | 81.4% | Human-in-the-loop | Jun 2025 | Official leaderboard |
| 2 | OpenAI Operator | 76.5% | Human-in-the-loop | May 2025 | Official leaderboard |
| 3 | Anthropic Sonnet 3.7 CUA | 66% | Autonomous | May 2025 | Official leaderboard |
| 4 | Skyvern 2.0 | 64.4% | Autonomous | May 2025 | Official leaderboard |
| 5 | Skyvern 2.0 on Browserbase | 60.7% | Autonomous | May 2025 | Official leaderboard |
| 6 | OpenAI Computer-Using Agent (CUA) | 59.8% | Autonomous | Jan 2025 | Official leaderboard |
| 7 | Browser Use Cloud | 43.9% | Autonomous | May 2025 | Official leaderboard |
| 8 | Convergence AI | 39.9% | Human-in-the-loop | May 2025 | Official leaderboard |
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official WebBench (Halluminate) leaderboard
No lab model card reports it yet
We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.
More in Computer and browser use
- BrowseComp In use Hard-to-find facts that need persistent web browsing. Short verifiable answers graded for exact correctness. Claude Fable 5 93.3%
- ScreenSpot-Pro In use GUI grounding on high-resolution professional-software screenshots. The model must locate the target UI element. GPT-6 Astra 92.7%
- AndroidWorld In use Agents complete 116 tasks across 20 real Android apps in an emulator. Each task checks the device's system state to judge success. GPT-6 Astra 100%
- OSWorld 2.1 In use Agents complete 108 long-horizon workflows across 31 self-hosted websites and desktop apps, averaging 318 tool calls. Scored by binary completion and checkpoint partial credit. Claude Opus 5.5 81.8%
- WebArena-Verified In use Agents complete 812 audited WebArena tasks on self-hosted websites. Deterministic evaluators check the agent's answer and captured network traces.
Muse Spark 1.1 69% - MobileWorld In use Agents complete 201 tasks across 20 mobile apps, including tasks that need user clarification or MCP tools. Backend checks verify success. GPT-5 55.3%