Cybench
Cybench is a benchmark for evaluating the cybersecurity capabilities and risks of language models. It includes 40 professional-level Capture the Flag (CTF) tasks from 4 distinct CTF competitions, chosen to be recent, meaningful, and spanning a wide range of difficulties.
Official Cybench Website
Paper:
Cybench: A Framework for Evaluating Cybersecurity Capabilities and Risks of Language Models (Zhang et al., 2024)
Key Features of Cybench
Professional CTF Tasks
Tasks sourced from real CTF competitions, representing actual cybersecurity challenges.
Diverse Categories
Tasks span various cybersecurity domains including web exploitation, cryptography, and reverse engineering.
Gradated Evaluation
Subtasks break down complex challenges into intermediary steps for more nuanced assessment.
Cybench Leaderboard
| Rank | Agent | Models |
Verified
Verified Results
Results have been reproduced by the HAL team |
Accuracy
Accuracy
Confidence intervals show the min-max values across runs for those agents where multiple runs are available |
Cost (USD)
Total Cost
Total API cost for running the agent on all tasks. Confidence intervals show the min-max values across runs for those agents where multiple runs are available |
Runs
Number of Runs
The number of runs for this agent submitted to the leaderboard. To submit multiple evaluations, rerun the same agent and set the same agent name |
Traces |
|---|---|---|---|---|---|---|---|
| 1 | o3-mini-2025-01-14 med. | ✓ | 25.00% | $5.88 | 1 | Download | |
| 2 | o1-preview-2024-09-12 | ✓ | 20.00% | $117.89 | 1 | Download | |
| 3 | claude-3-5-sonnet-20241022 | ✓ | 20.00% | $12.90 | 1 | Download | |
| 4 | gpt-4.5-preview-2025-02-27 | ✓ | 17.50% | $265.89 | 1 | Download | |
| 5 | claude-3-7-sonnet-20250219 | ✓ | 17.50% (-0.00/+0.00) | $16.95 (-0.92/+0.92) | 2 | Download | |
| 6 | o1-mini-2024-09-12 | ✓ | 12.50% | $10.44 | 1 | Download | |
| 7 | gpt-4o-2024-11-20 | ✓ | 12.50% | $7.97 | 1 | Download | |
| 8 | gpt-4o-mini-2024-07-18 | ✓ | 7.50% (-0.00/+0.00) | $0.48 (-0.02/+0.02) | 2 | Download | |
| 9 | Meta-Llama-3.1-405B-Instruct-Turbo | ✓ | 0.00% | $2.93 | 1 | Download |
Accuracy vs. Cost Frontier for Cybench
This plot shows the relationship between an agent's performance and its token cost. The Pareto frontier (dashed line) represents the current state-of-the-art trade-off. The error bars indicate min-max values across runs.
Heatmap for CyBench
The heatmap visualizes success rates across tasks and agents. Colorscale shows the fraction of times a task was solved across reruns of the same agent. The "any agent" performance indicates the level of saturation of the benchmark and gives a sense of overall progress.
Failure Analysis (Experimental)
Select an agent to see a detailed breakdown of failure categories and their descriptions. This analysis helps understand common failure patterns and areas for improvement. Failure reports are usually available for the top 2 agents.
Failure Categories
Distribution of Failures
Token Pricing Configuration
Adjust token prices to see how they affect the total cost calculations in the leaderboard and plots.
Additional Resources
Getting Started
Want to evaluate your agent on Cybench? Follow our comprehensive guide to get started:
View Documentation