Clawbench: Open-Source Benchmark for Browser AI Agents

Clawbench is an open-source benchmark for evaluating AI agents on daily tasks performed in a web browser. It was created by TIGER-AI-Lab and last updated on May 25, 2026. The benchmark likely contains tasks designed to test an agent's ability to interact with web interfaces.

Use Cases

• Benchmarking AI agent performance on daily web tasks based on the described evaluation framework.
• Training browser-based AI assistants on structured tasks mentioned in the description.
• Comparing different AI agent architectures on a standardized set of web interaction challenges.

Strengths

• Released under the permissive Apache-2.0 license, allowing for broad use and modification.
• Open-source nature facilitates community review and contribution to the benchmark.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for large-scale training.

Provenance

Source

TIGER-AI-Lab

Freshness

Last updated 2026-05-25 00:33:10; freshness should be verified.