Benchmark Almanac

CyberGym-E2E

Claude Opus 5 89% leads since Jul 2026 Not in model cards

Agents get only source code and must find a real vulnerability, write a proof-of-concept input and a patch that keeps functionality. Covers 920 vulnerabilities in 139 open-source projects.

920 instances from 139 open-source projects. Agents discover a vulnerability, write a proof of concept and patch it. Four cumulative stages S1 to S4, plus a patch-only setting. Some rows use an earlier 615-task set.

In the authors’ words

AI has the potential to transform cybersecurity by enabling systems that can autonomously detect, analyze, and remediate software vulnerabilities. However, existing cybersecurity evaluations of AI systems are limited in scale or scope, and fail to capture the end-to-end lifecycle of real-world software vulnerability discovery and remediation. To address this gap, we propose CyberGym-E2E, a large-scale and realistic end-to-end cybersecurity benchmark that comprehensively evaluates AI agents' abilities across the full lifecycle of vulnerability discovery, PoC generation, and patch generation. CyberGym-E2E is comprehensive and scalable, as we build an automated, agent-enhanced pipeline for transforming open-source vulnerability data into realistic evaluation environments. Currently, the benchmark consists of 920 real-world vulnerabilities across 139 different open-source projects.

Abstract of CyberGym-E2E: Scalable Real-World Benchmark for AI Agents' End-to-End Cybersecurity Capabilities (arXiv:2606.04460, CC0)
Made by UC Berkeley RDI (Sunblaze) and collaborators Released Jun 2026 Finding and exploiting real bugs All tasks public Last model added Jul 2026 Scores refreshed 2 Nov 2026
Level
Expert Specialists with years of training, often over hours of work
Scored
Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
Answers
Agent task Many steps in a terminal, browser, desktop, app, game or robot.
sunblaze-ucb/cybergym-e2e 92 stars Paper Leaderboard cybergym.io
5
Models scored on its own leaderboard

Claude Opus 5 leads with 89% among models released by Jul 2026

Best score per model from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best.

The newest model with a score here came out in Jul 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 89.0 Claude Opus 5
  2. 65.9 GPT-5.4
  3. 62.6 Claude Opus 4.6
  4. 43.8 Gemini 3.1 Pro
  5. 28.4 DeepSeek V4 Flash 0731
  1. 1 Claude Opus 5 89%
  2. 2 GPT-5.4 65.9%
  3. 3 Claude Opus 4.6 62.6%
  4. 4 Gemini 3.1 Pro 43.8%
  5. 5 DeepSeek V4 Flash 0731 28.4%
#ModelScoreSettingDateSource
1 Claude Opus 5 89% AWS Continuum Jul 2026 Official leaderboard
2 GPT-5.4 65.9% Codex Mar 2026 Official leaderboard
3 Claude Opus 4.6 62.6% Claude Code Feb 2026 Official leaderboard
4 Gemini 3.1 Pro 43.8% Gemini CLI Feb 2026 Official leaderboard
5 DeepSeek V4 Flash 0731 28.4% Aegis Jul 2026 Official leaderboard

2 models have held number one since its release in Jun 2026

Lab claims count from the day the card came out, independent runs from the model’s release.

OpenAI Anthropic
Own leaderboard Passed 85% in Jul 2026, by Claude Opus 5
Claude Opus 5 89%

The best score passed 85% in Jul 2026

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Claude Opus 5 GPT-5.4 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Claude Opus 5 GPT-5.4

No lab model card reports it yet

We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark.

More in Cybersecurity

All 111 in Cybersecurity
  • ExploitGym In use 869 real-vulnerability exploit-development tasks (userspace, V8, Linux kernel). Success means capturing a flag through the intended vulnerability. GPT-6 Astra 42.4%
  • ExploitBench In use 41 recent V8 JavaScript-engine N-day vulnerabilities. Progress measured by 16 grader-verified capability flags across five tiers. GPT-6 Astra 100%
  • CWE-bench In use Agents audit a repository and write a patch for a security flaw, without being told where the bug is. Covers six programming languages. Gemini 4 Carbon 71%
  • Real-world vulnerability discovery In use Agents search popular open-source codebases to rediscover about 1,200 recent confirmed vulnerabilities across many languages. Reports share discovered. Gemini 4 Argon 86%
  • Artificial Analysis Cyber Index In use Agents do enterprise cyber defense tasks from three benchmarks, CWE-Bench-AA, DeepsecBench-AA and CyberGym-E2E-AA, combined into one success rate. Mistral Large 4 (Le Chonk, open weights) 51%
  • Big Sleep Evaluation (Google) In use Agents search large real codebases such as Chrome and Safari for critical, hard to find vulnerabilities. Scored by pass@1 success rate, run without safety guardrails. Gemini 3.5 Flash Cyber 72%