HackSynth PicoCTF
Models attempt 120 tasks selected from picoCTF. The benchmark measures the percentage of tasks completed, and the hub port checks submitted answers against reference flags.
In December 2024, GPT-4o led eight models at 34.2%. The comparison included GPT-4o mini, Llama, Mixtral, Qwen, and Phi.
- Made by
- Eötvös Loránd University
- Released
- December 2024
- Kind
- Capture-the-flag challenges
- Level
- Competition
- Scored
- Checked automatically
- Answers
- Agent task
- Tasks
- All public
- Measures
- Capability
- Risk
- Cyber offence
- Language
- English
- Interaction
- Agentic
- Environment
- Terminal
- Items from
- Exams and contests, Another benchmark
- Refresh
- One fixed set
- Run by
- Anyone
- Based on
- A version of another benchmark
- In views
- Agents, Dangerous capabilities
- Strongest model tested
-
Llama 3.1-70B #177 on the capability index
- Last model added
- August 2024
- Scores refreshed
- November 20th, 2026
GPT-4o (snapshot unspecified) tops the leaderboard with 34.2% Llama 3.1-70B leads with 22.5% among models released by August 2024
Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. 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 best model we track is Llama 3.1-70B, at 22.5%. This view shows only models we track, each at its best.
The newest model with a score here came out in August 2024, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 1 GPT-4o (snapshot unspecified) 34.2%
- 2 GPT-4o-mini (snapshot unspecified) 24.2%
- 3
Llama 3.1-70B 22.5%
- 4
Qwen2-72B 20.8%
- 5
Llama 3.1-8B 15.8%
- 6
Mixtral 8x7B 15%
- 7
Phi-3.5-MoE 11.7%
- 8 Phi-3-mini 5.00%
Source: The HackSynth PicoCTF paper
- 1
Llama 3.1-70B 22.5%
- 2
Qwen2-72B 20.8%
- 3
Llama 3.1-8B 15.8%
- 4
Mixtral 8x7B 15%
- 5
Phi-3.5-MoE 11.7%
Source: The HackSynth PicoCTF paper
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-4o (snapshot unspecified) | 34.2% | HackSynth | Official leaderboard | |
| 2 | GPT-4o-mini (snapshot unspecified) | 24.2% | HackSynth | Official leaderboard | |
| 3 | | 22.5% | HackSynth | Jul 2024 | Official leaderboard |
| 4 | | 20.8% | HackSynth | Jun 2024 | Official leaderboard |
| 5 | | 15.8% | HackSynth | Jul 2024 | Official leaderboard |
| 6 | | 15% | HackSynth | Dec 2023 | Official leaderboard |
| 7 | | 11.7% | HackSynth | Aug 2024 | Official leaderboard |
| 8 | Phi-3-mini | 5.00% | HackSynth | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 22.5% | HackSynth | Jul 2024 | Official leaderboard |
| 2 | | 20.8% | HackSynth | Jun 2024 | Official leaderboard |
| 3 | | 15.8% | HackSynth | Jul 2024 | Official leaderboard |
| 4 | | 15% | HackSynth | Dec 2023 | Official leaderboard |
| 5 | | 11.7% | HackSynth | Aug 2024 | Official leaderboard |
Rankings by part
We read 6 parts from the benchmark’s paper, each ranking models on one part of the test, such as General (34), Forensics (21) and Crypto (23). Qwen2-72B leads 3 of them, and Mixtral 8x7B leads the other 3. Arrows show how far a model moved from its overall rank among the models we track.
General (34)
5 models we track have a score here.
- 1
Qwen2-72B ↑1 44.1%
- 2
Llama 3.1-70B ↓1 44.1%
- 3
Mixtral 8x7B ↑1 26.5%
- 4
Llama 3.1-8B ↓1 26.5%
- 5
Phi-3.5-MoE 26.5%
Forensics (21)
5 models we track have a score here.
- 1
Mixtral 8x7B ↑3 19%
- 2
Llama 3.1-70B ↓1 19%
- 3
Llama 3.1-8B 19%
- 4
Qwen2-72B ↓2 9.50%
- 5
Phi-3.5-MoE 4.80%
Crypto (23)
5 models we track have a score here.
- 1
Mixtral 8x7B ↑3 13%
- 2
Qwen2-72B 13%
- 3
Llama 3.1-70B ↓2 13%
- 4
Phi-3.5-MoE ↑1 13%
- 5
Llama 3.1-8B ↓2 8.70%
Web (21)
5 models we track have a score here.
- 1
Qwen2-72B ↑1 9.50%
- 2
Llama 3.1-70B ↓1 9.50%
- 3
Llama 3.1-8B 4.80%
- 4
Mixtral 8x7B 0%
- 5
Phi-3.5-MoE 0%
Rev (19)
5 models we track have a score here.
- 1
Qwen2-72B ↑1 15.8%
- 2
Llama 3.1-70B ↓1 15.8%
- 3
Llama 3.1-8B 15.8%
- 4
Mixtral 8x7B 10.5%
- 5
Phi-3.5-MoE 5.30%
Bin (2)
5 models we track have a score here.
- 1
Mixtral 8x7B ↑3 0%
- 2
Qwen2-72B 0%
- 3
Llama 3.1-70B ↓2 0%
- 4
Llama 3.1-8B ↓1 0%
- 5
Phi-3.5-MoE 0%
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: The HackSynth PicoCTF paper
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official HackSynth PicoCTF leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 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%
- ExploitBench - Internal Port In use Agents develop working exploits for recently disclosed vulnerabilities to limit contamination. Scored as arbitrary code execution success rate. GPT-6.2 Astra 53.5%
- Atomic Challenges In use Agents solve 22 skill-level offensive security challenges in vulnerability research, network attack simulation and evasion. Scored as challenges solved. GPT-6 Astra 90.9%
- MiMo Cyber Bench In use Agents do in-house cybersecurity tasks that add to public vulnerability reproduction and exploitation benchmarks. The report gives no task count or grading details. MiMo-V2.6 Pro 81.7%
- PatchBench (OpenAI) In use Models patch security flaws. Automated tests and model-based review check that each patch fixes the flaw, keeps the intended behavior and is ready to merge. GPT-6.2 Astra 79.7%
Data audit
9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 2 years ago, in August 2024, for Phi-3.5-MoE. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
- Pass Independent runs Scores from the benchmark’s paper (8 entries).
- Pass Models scored 5 models we track, plus 3 other entries on its leaderboard.
- Warning Newest score The newest came 2 years ago, in August 2024, for Phi-3.5-MoE.
- Note Frontier models None of the top 10 has a score. The strongest tested is Llama 3.1-70B (#177).
- Pass Headroom The best is 22.5%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on arxiv.org: 5 rows for models we track and 3 other entries at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.