README.md

metadata

license: apache-2.0
task_categories:
  - text-generation
  - question-answering
  - other
language:
  - en
tags:
  - benchmark
  - leaderboard
  - agent-benchmark
  - llm-benchmark
  - web-agents
  - browser-agent
  - browser-automation
  - ai-agent
  - evaluation
  - real-world-tasks
  - web-navigation
  - task-completion
  - clawbench
  - multimodal
pretty_name: 'ClawBench: Web Agent Benchmark'
size_categories:
  - n<1K
configs:
  - config_name: default
    data_files:
      - split: test
        path: data/train-00000-of-00001.parquet
arxiv: '2604.08523'
viewer: true
leaderboard: NAIL-Group/clawbench-leaderboard

ClawBench — A Benchmark for AI Web Agents

Can AI Agents Complete Everyday Online Tasks?

|💻 Github | 🏆 Leaderboard | 📖 Paper | 🌐 Website |

ClawBench is an open benchmark for AI web agents — the systems that drive a real browser to complete a user's task end-to-end. It scores agents on real, everyday online tasks (booking flights, ordering groceries, submitting job applications) across live websites. The corpus ships in two slices: V1 — 153 tasks across 144 websites (the original frontier-model leaderboard) and V2 — 130 newer tasks (expanded coverage). For each run we capture 5 layers of behavioral data (session replay, screenshots, HTTP traffic, agent reasoning traces, and browser actions), collect human ground-truth, and score with an agentic evaluator that provides step-level traceable diagnostics.

Install: pip install clawbench-eval (PyPI) · Companion raw traces: NAIL-Group/ClawBenchV1Trace

🏆 Leaderboard

Live results — pulled from leaderboard/results.csv in this repo. Sort by corpus (v1 / v2 / all) and submit your model in the interactive Space:

Snapshot — last refreshed 2026-05-10:

Rank	Model	Harness	Corpus	Pass	Total	Pass Rate	Wall (h)
1	glm-5.1	hermes	v2	63	130	48.46%	11.35
2	glm-5.1	hermes	v1	25	153	16.34%	15.37
3	openrouter/owl-alpha	hermes	v2	19	130	14.62%	7.58
4	deepseek/deepseek-v4-flash	hermes	v1	14	153	9.15%	13.59
5	glm-5.1	openclaw	v1	13	153	8.50%	6.50
6	deepseek/deepseek-v4-flash	hermes	v2	4	130	3.08%	2.37
7	poolside/laguna-m.1:free	hermes	v1	1	153	0.65%	1.52

Submit a result → run clawbench-eval on your model and open a PR to leaderboard/results.csv — one row per (model × harness × corpus).

Companion dataset (raw traces): NAIL-Group/ClawBenchV1Trace — recording.mp4, requests.jsonl, actions.jsonl, agent-messages.jsonl, interception.json, run-meta.json per V1 model run.

Dataset Structure

Columns

Column	Type	Description
task_id	int	Unique task identifier
instruction	string	Task prompt sent to the agent
metaclass	string	High-level category (21 categories)
class	string	Fine-grained sub-category
platform	string	Target platform (144 unique platforms)
sites	list[string]	Domains involved in the task
eval_schema	string (JSON)	Request interception configuration
time_limit	int	Maximum time in minutes
extra_info	string (JSON)	Paths to additional context files
shared_info	string	Path to shared user profile

Additional Files

shared/
  alex_green_personal_info.json   # Shared dummy user profile used across all tasks
extra_info/
  004/grocery_list.json           # Task-specific context (32 tasks have extra info)
  007/meal_plan.json
  043/pet_info.json
  ...

• shared/alex_green_personal_info.json — A comprehensive dummy user persona (Alex Green) including personal details, address, work history, education, financial information, and preferences. All tasks share this identity.
• extra_info/ — Task-specific supplementary files referenced by the extra_info column. 32 of 153 tasks include additional context such as grocery lists, job links, meeting details, etc.

eval_schema

The eval_schema field configures the request interceptor — a mechanism that blocks the final HTTP request matching the specified URL pattern and method, preventing irreversible actions (checkout, form submission, etc.) from reaching the server. This allows safe evaluation on live websites.

{
  "url_pattern": "taskrabbit\\.(com|ca)/(api/v\\d+/jobs|book/\\d+/confirm)",
  "method": "POST"
}

Task Categories (metaclass)

Category	Tasks	Example Platforms
daily-life	21	Uber Eats, Instacart, Zillow
entertainment-hobbies	15	Goodreads, Eventbrite, Fandango
creation-init	13	ClickUp, Typeform, Ghost
office-secretary-tasks	9	Trello, Calendly, Purelymail
rating-voting	10	TripAdvisor, Glassdoor, Yelp
education-learning	9	Coursera, LeetCode, Blinkist
travel	9	Google Flights, Hipcamp, Airbnb
beauty-personal-care	9	TaskRabbit, Booksy, Soko Glam
pet-animal-care	8	Rover, Petfinder, Chewy
job-search-hr	8	Indeed, Greenhouse, ZipRecruiter
academia-research	5	Zotero, Overleaf, Google Scholar
and 10 more...

Usage

from datasets import load_dataset

ds = load_dataset("NAIL-Group/ClawBench", split="test")
print(ds[0])

Citation

@article{zhang2026clawbench,
  title={ClawBench: Can AI Agents Complete Everyday Online Tasks?},
  author={Yuxuan Zhang and Yubo Wang and Yipeng Zhu and Penghui Du and Junwen Miao and Xuan Lu and Wendong Xu and Yunzhuo Hao and Songcheng Cai and Xiaochen Wang and Huaisong Zhang and Xian Wu and Yi Lu and Minyi Lei and Kai Zou and Huifeng Yin and Ping Nie and Liang Chen and Dongfu Jiang and Wenhu Chen and Kelsey R. Allen},
  journal={arXiv preprint arXiv:2604.08523},
  year={2026}
}