Wiz Cyber Model Arena (ADK React) - Code Vulnerabilities: leaderboard
Wiz Cyber Model Arena scores AI agents on offensive security challenges in four categories (code vulnerabilities, API security, Websec.fr CTFs, cloud security). This board is the percentage of code vulnerability challenges solved on the first attempt (pass@1) by models run in the ADK React harness; higher is better.
Metric: Code Vulnerabilities challenges solved, pass@1 (%). Source: www.wiz.io. Saturation forecast: Around July 2027. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 4 Argon | 65.8 | #9 |
| 2 | Claude Opus 5.5 | 58.7 | #3 |
| 3 | GPT-5.6 Sol | 41.8 | #22 |
| 4 | Claude Opus 4.8 | 36.3 | #39 |
| 5 | GPT-5.6 Terra | 35.3 | #40 |
| 6 | DeepSeek V4.1 Flash (1017) | 34.7 | #27 |
| 7 | Claude Haiku 5.5 | 34.1 | #43 |
| 8 | Claude Opus 5 | 31.4 | #16 |
| 9 | DeepSeek V4 Pro | 27.5 | #112 |
| 10 | Kimi K3 | 27.5 | #35 |
| 11 | Gemini 3.6 Flash | 25.8 | #57 |
| 12 | Grok 4.6 | 23.4 | #29 |
| 13 | GLM-5.3 | 21.7 | #44 |
| 14 | Claude Sonnet 4.5 | 19 | #148 |
| 15 | Gemini 3.5 Flash | 17.3 | #66 |
No result here: #6 GPT-6 Astra, #7 GPT-6.1 Sol, #10 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=wiz-cyber-model-arena-adk-react-code-vulnerabilities · How It Works · Data refreshed daily, snapshot 2026-11-05.