CyberGym-E2E
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.
- Made by
- UC Berkeley RDI (Sunblaze) and collaborators
- Released
- June 2026
- Kind
- Finding and exploiting real bugs
- Level
- Expert
- Scored
- Checked automatically
- Answers
- Agent task
- Tasks
- All public
- Strongest model tested
- Claude Opus 5 #10 on the capability index
- Last model added
- July 2026
- Scores refreshed
- November 12th, 2026
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.
Claude Opus 5 leads with 89% among models released by July 2026
Best score per entry from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.
The newest model with a score here came out in July 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- Claude Fable 5.2
-
Gemini 4 Carbon - GPT-6 Astra
- Claude Opus 5.5
- and 15 more
- 1 Claude Opus 5 No cost cap / 90 min 89%
- 2 Multi-model (DeepSeek-V4-Flash, Zhian-27B) self-hosted / 6 h 66.5%
- 3 GPT-5.4 $10 / 90 min 65.9%
- 4 Claude Opus 4.6 No cost cap / 90 min 62.6%
- 5 Gemini 3.1 Pro $10 / 90 min 43.8%
- 6
DeepSeek V4 Flash 0731 self-hosted / 90 min 28.4%
Source: Official CyberGym-E2E leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Opus 5 | 89% | No cost cap / 90 min, AWS Continuum | Jul 2026 | Official leaderboard |
| 2 | Multi-model (DeepSeek-V4-Flash, Zhian-27B) | 66.5% | self-hosted / 6 h, Zhian·Changying (智安·长缨) | Oct 2026 | Official leaderboard |
| 3 | GPT-5.4 | 65.9% | $10 / 90 min, Codex | Mar 2026 | Official leaderboard |
| 4 | Claude Opus 4.6 | 62.6% | No cost cap / 90 min, Claude Code | Feb 2026 | Official leaderboard |
| 5 | Gemini 3.1 Pro | 43.8% | $10 / 90 min, Gemini CLI | Feb 2026 | Official leaderboard |
| 6 | DeepSeek V4 Flash 0731 | 28.4% | self-hosted / 90 min, Aegis | Jul 2026 | Official leaderboard |
2 models have held number one on the official leaderboard since its release in June 2026
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official CyberGym-E2E leaderboard
Claude Opus 5 on the official leaderboard passed 85% in July 2026
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official CyberGym-E2E leaderboard
No lab model card reports it yet
We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.
More 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%
- CyberBench v1.1 (Vals) In use Agents work on OSS-Fuzz memory-safety bugs from ARVO images in two tracks. They write a proof-of-concept input that crashes the target, or patch the source so the crash stops and tests still pass.
Muse Wonder 80.4% - SRE-Bench In use Software reverse engineering without source: 262 binary instances from 19 private programs, each with six graded objectives. GPT-6 Astra 56.9%
- CyScenarioBench In use Scenario-based long-horizon security challenges in sandboxes, organized in task, scenario and campaign tiers. Evaluation set kept private. Gemini 4 Argon 100%
- ExploitBench - Internal Port In use Agents develop working exploits for recently disclosed vulnerabilities to limit contamination. Scored as arbitrary code execution success rate. GPT-6 Cyber 42%