Cybench

Benchmark / dataset

40 professional CTF tasks with subtasks; used in Anthropic, OpenAI and AISI system cards.

Category: Cyber offense capability
Insurance exposure: Breach/cyber
Publisher: Stanford
Publisher type: Academic
License: Apache-2.0
Status: Active (<90d)
Last commit: 2026-09-24
Stars: 342
Notes: Industry-standard offensive cyber eval.

View source on GitHub →

Wanich Risk Lab, 2026.