In-house CTF suite (Phuong et al. 2024)
Gemini 2.5 Pro Jun 2025 84.6% leads since July 2025 Agents hack 13 simulated servers with vulnerable web apps and services to retrieve a secret flag.
Challenges cover web app vulnerabilities, off-the-shelf exploits, databases, Linux privilege escalation and password cracking and spraying. Challenge files without flags or solutions are in the dangerous-capability-evaluations repo. UK AISI inspect_evals ships it as gdm_in_house_ctf, revised seven times to September 2026.
- Made by
- Google DeepMind
- Released
- March 2024
- Kind
- Capture-the-flag challenges
- Level
- Expert
- Scored
- Checked automatically
- Answers
- Agent task
- Tasks
- All public
- Strongest model tested
- Gemini 2.5 Pro Jun 2025 #96 on the capability index
- Last in a system card
- July 2025
- Last model added
- July 2025
Gemini 2.5 Pro leads with 84.6% among models released by June 2025
Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.
The newest model with a score here came out in June 2025, and nobody has reported 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 16 more
- 1 Gemini 2.5 Pro Jun 2025 84.6%
- 2 Gemini 2.5 Pro Mar 2025 76.9%
- 3 Gemini 2.5 Flash Jun 2025 76.9%
- 4 Gemma 2 27B 7.70%
Source: Labs’ model cards
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Gemini 2.5 Pro Jun 2025 | 84.6% | Jul 2025 | Lab model card | |
| 2 | Gemini 2.5 Pro Mar 2025 | 76.9% | Apr 2025 | Lab model card | |
| 3 | Gemini 2.5 Flash Jun 2025 | 76.9% | Jul 2025 | Lab model card | |
| 4 | Gemma 2 27B | 7.70% | Jul 2024 | Lab model card |
3 models have held number one in labs’ own numbers since its release in March 2024
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Labs’ model cards
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Labs’ model cards
Labs report it only in system cards
We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark. Some system cards do, below.
4 system cards report it, from 1 lab
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| Jul 7, 2025 | Gemini 2.5 Pro technical report | 84.6 | In-house CTF (medium), 11/13 solved (Table 10, Figure 10) | Cyber risk |
| Jul 7, 2025 | Gemini 2.5 Flash technical report | 76.9 | In-house CTF (medium), 10/13 solved (Figure 10 bar label) | Cyber risk |
| Apr 28, 2025 | Gemini 2.5 Pro model card | 76.9 | In-house CTF (medium), 10/13 solved (Table 1, Figure 2) | Cyber risk |
| Jul 31, 2024 | Gemma 2 27B technical report | 7.7 | Internal CTF suite, 1/13 challenges solved (8%) (Table 19) | Cyber risk |
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%