Benchmark Almanac

Capture the Flag (Professional)

GPT-5.5 96.3% leads since Apr 2026 Fading

Agents solve curated public professional-level CTF challenges in a Linux box with security tools. Scored as pass@12 over 16 rollouts.

OpenAI Preparedness set built from public CTFs. OpenAI refactored the set in July 2025.

Made by OpenAI Released Aug 2024 Capture-the-flag challenges Tasks kept private Last in a launch post Dec 2025 , dropped after it saturated Last model added May 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.
21
Models scored in lab cards
1
Launch post lists it, from 1 lab
22
System cards report it, from 1 lab

GPT-5.5 leads with 96.3% among models released by May 2026

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 May 2026, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

21 models. Scroll sideways for the rest.

  1. 96.3 GPT-5.5
  2. 94.1 GPT-5.5 Instant
  3. 88.2 GPT-5.4
  4. 88.1 GPT-5.2 Codex
  5. 88.0 GPT-5.3 Codex
  6. 82.0 GPT-5.2
  7. 81.3 GPT-5.4 Mini
  8. 76.0 GPT-5.1-Codex-Max
  9. 55.0 o3
  10. 47.0 OpenAI deep research
  11. 41.0 o4-mini
  12. 27.0 GPT-5
  13. 23.0 GPT-5 mini
  14. 21.0 o3-mini
  15. 14.0 gpt-oss-120b
  16. 13.0 o1
  17. 7.30 o1-preview
  18. 3.90 o1-mini
  19. 2.00 GPT-4.5
  20. 2.00 gpt-oss-20b
  21. 1.00 GPT-4o
  1. 1 GPT-5.5 96.3%
  2. 2 GPT-5.5 Instant 94.1%
  3. 3 GPT-5.4 88.2%
  4. 4 GPT-5.2 Codex 88.1%
  5. 5 GPT-5.3 Codex 88%
  6. 6 GPT-5.2 82%
  7. 7 GPT-5.4 Mini 81.3%
  8. 8 GPT-5.1-Codex-Max 76%
  9. 9 o3 55%
  10. 10 OpenAI deep research 47%
  11. 11 o4-mini 41%
  12. 12 GPT-5 27%
  13. 13 GPT-5 mini 23%
  14. 14 o3-mini 21%
  15. 15 gpt-oss-120b 14%
  16. 16 o1 13%
  17. 17 o1-preview 7.30%
  18. 18 o1-mini 3.90%
  19. 19 GPT-4.5 2.00%
  20. 20 gpt-oss-20b 2.00%
  21. 21 GPT-4o May 2024 1.00%
#ModelScoreSettingDateSource
1 GPT-5.5 96.3% Apr 2026 Model card
2 GPT-5.5 Instant 94.1% May 2026 Model card
3 GPT-5.4 88.2% Mar 2026 Model card
4 GPT-5.2 Codex 88.1% Dec 2025 Model card
5 GPT-5.3 Codex 88% Feb 2026 Model card
6 GPT-5.2 82% Dec 2025 Model card
7 GPT-5.4 Mini 81.3% Mar 2026 Model card
8 GPT-5.1-Codex-Max 76% Nov 2025 Model card
9 o3 55% Apr 2025 Model card
10 OpenAI deep research 47% Feb 2025 Model card
11 o4-mini 41% Apr 2025 Model card
12 GPT-5 27% Aug 2025 Model card
13 GPT-5 mini 23% Aug 2025 Model card
14 o3-mini 21% Jan 2025 Model card
15 gpt-oss-120b 14% Aug 2025 Model card
16 o1 13% Dec 2024 Model card
17 o1-preview 7.30% Sep 2024 Model card
18 o1-mini 3.90% Sep 2024 Model card
19 GPT-4.5 2.00% Feb 2025 Model card
20 gpt-oss-20b 2.00% Aug 2025 Model card
21 GPT-4o May 2024 1.00% Aug 2024 Model card

11 models have held number one since its release in Aug 2024

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

OpenAI
Lab claims Passed 85% in Dec 2025, by GPT-5.2 Codex
GPT-5.5 96.3%

The best score passed 85% in Dec 2025

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

Reported by the lab in its model card Best so far
0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) GPT-5.5 GPT-5.4 GPT-5.2 GPT-5.1-Codex-Max o3 OpenAI deep research o3-mini 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) GPT-5.5 GPT-5.4

Labs headline it less often now, last in Dec 2025

Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.

OpenAI
DateLaunch postScoreAs the post writes it
18 Dec 2025 GPT-5.2 Codex 88.1 Professional Capture-the-Flag (CTF) eval

22 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.

DateDocumentScoreAs the document writes itSection
4 May 2026 GPT-5.5 Instant system card 94.1 Capture the Flag (Professional), pass@12, run above deployed reasoning effort, printed chart label (Figure 7, 8.1.2.1) Cyber risk
23 Apr 2026 GPT-5.5 system card 96.3 Capture the Flag (Professional), pass@12, printed chart label (Figure 16, 9.1.2.1) Cyber risk
17 Mar 2026 GPT-5.4 Mini system card 81.3 Capture the Flag (CTF), pass@12 (Table 19) Cyber risk
5 Mar 2026 GPT-5.4 system card 88.2 Capture the Flag (Professional), pass@12, printed chart label (Figure 12, 5.1.2.1) Cyber risk
5 Feb 2026 GPT-5.3 Codex system card 88 Professional CTFs, pass@12, printed chart label (Figure 5, 5.1.2.1) Cyber risk
18 Dec 2025 GPT-5.2 Codex system card addendum 88 Professional CTFs, pass@12, printed chart label (Figure 5, 5.1.2.1) Cyber risk
11 Dec 2025 GPT-5.2 system card addendum 82 Professional CTFs, pass@12, gpt-5.2 no browsing, printed chart label (Figure 9) Cyber risk
18 Nov 2025 GPT-5.1-Codex-Max system card 76 Professional CTFs, pass@12, printed chart label (Figure 6, 5.1.2.1) Cyber risk
7 Aug 2025 GPT-5 system card 27 Professional CTFs refactored, pass@12, no browsing launch candidate (Figure 14) Cyber risk
7 Aug 2025 GPT-5 mini system card 23 Professional CTFs refactored, pass@12, gpt-5-thinking-mini no browsing launch candidate (Figure 14) Cyber risk
5 Aug 2025 gpt-oss-20b model card 2 Capture the Flag (CTF) Professional, pass@12, gpt-oss 20b no browsing launch candidate (Figure 11) Cyber risk
5 Aug 2025 gpt-oss-120b model card 14 Capture the Flag (CTF) Professional, pass@12, gpt-oss 120b no browsing launch candidate (Figure 11) Cyber risk
17 Jul 2025 ChatGPT agent system card Professional CTFs refactored, pass@12, chart shows only ChatGPT Agent no-browsing helpful-only at 23%, no released-model value (Figure 5, 5.1.2.1) Cyber risk
16 Apr 2025 o4-mini system card 41 Professional CTFs, pass@12, o4-mini no browsing launch candidate, printed chart label (Figure 7), text summary gives 41% Cyber risk
16 Apr 2025 o3 system card 55 Professional CTFs, pass@12, o3 no browsing launch candidate, printed chart label (Figure 7), text summary gives 59% Cyber risk
27 Feb 2025 GPT-4.5 system card 2 Professional CTFs, pass@12, post-mitigation (4.2) Cyber risk
25 Feb 2025 OpenAI deep research system card 47 Professional CTFs, pass@12, post-mitigation without browsing, text and printed chart label (with browsing 70%) (3.4.3) Cyber risk
31 Jan 2025 o3-mini system card 21 Professional CTFs, pass@12, post-mitigation (5.3) Cyber risk
5 Dec 2024 o1 system card 13 Professional CTFs, pass@12, o1 (post-mitigation) (5.4) Cyber risk
12 Sep 2024 o1-preview system card 7.3 Professional CTFs, pass@12, final o1-preview post-mitigation model (near-final checkpoint 26.7%, 0%, 2.5%) (4.2) Cyber risk
12 Sep 2024 o1-mini system card 3.9 Professional CTFs, pass@12, o1-mini (post-mitigation) (4.2) Cyber risk
8 Aug 2024 GPT-4o system card 1 Capture the Flag, professional level, 10 attempts per task (3.5) Cyber risk

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%