Atlas
Benchmarks
331 sources · 1,258 models × 2,980 benchmarks · 155,643 scoresWiz Cyber Model Arena v3.9 (API Security)
The API Security category of Wiz's Cyber Model Arena v3.9: 54 black-box labs in which an agent discovers and exploits web vulnerabilities through a live API, with no source code and often no API specification, and must reach every exploitable path. Pass@1 averages three attempts per lab, for a model in a general-purpose harness.
Models
24
Top score
87.0
Leader
Gemini 4 Argon
Updated
—
Resolution
—
Saturation
—
No signal or difficulty estimate. Held out of the fit by design.
Organizations
| # | Model | Setting | Score | Source |
|---|---|---|---|---|
| 1 | Gemini 4 Argon | high · Antigravity CLI | 87.0 | |
| 2 | Gemini 3.8 Flash Cyber | high · Wiz Cyber Model Arena ADK React; simple ReAct loop built with ADK; single bash tool; no cyber-specific tools; general system prompt | 85.2 | |
| 3 | Claude Opus 5 | high · Claude Code | 84.0 | |
| 4 | DeepSeek V4.1 Flash 1017 | high · Claude Code | 83.3 | |
| 5 | GLM-5.3 | high · Claude Code | 82.7 | |
| 6 | GPT-5.6 Sol | high · Claude Code | 82.7 | |
| 7 | Gemini 3.7 Flash | high · Claude Code | 82.1 | |
| 8 | GPT-6 Astra | high · Claude Code | 81.5 | |
| 9 | DeepSeek V4.1 Flash | high · Wiz Cyber Model Arena ADK React; simple ReAct loop built with ADK; single bash tool; no cyber-specific tools; general system prompt | 80.9 | |
| 10 | GPT-6.1 Sol | high · Codex | 80.9 | |
| 11 | Gemini 3.6 Flash | high · Wiz Cyber Model Arena ADK React; simple ReAct loop built with ADK; single bash tool; no cyber-specific tools; general system prompt | 80.2 | |
| 12 | Opus 4.8 | high · Claude Code | 79.0 | |
| 13 | GLM-5.3 Flash | high · Wiz Cyber Model Arena ADK React; simple ReAct loop built with ADK; single bash tool; no cyber-specific tools; general system prompt | 78.4 | |
| 14 | GPT-5.6 Terra | high · Claude Code | 78.4 | |
| 15 | GPT-6 Sol | high · Claude Code | 78.4 | |
| 16 | Claude Haiku 5.5 | high · Claude Code | 77.8 | |
| 17 | MiMo-V2.6-Pro | high · Wiz Cyber Model Arena ADK React; simple ReAct loop built with ADK; single bash tool; no cyber-specific tools; general system prompt | 77.8 | |
| 18 | Sonnet 5 | high · Wiz Cyber Model Arena ADK React; simple ReAct loop built with ADK; single bash tool; no cyber-specific tools; general system prompt | 77.2 | |
| 19 | DeepSeek V4 Pro 0813 | high · Wiz Cyber Model Arena ADK React; simple ReAct loop built with ADK; single bash tool; no cyber-specific tools; general system prompt | 77.2 | |
| 20 | Kimi K3 | high · Claude Code | 75.9 | |
| 21 | GPT-6 Luna | high · Claude Code | 75.3 | |
| 22 | Gemini 3.5 Flash | high · Wiz Cyber Model Arena ADK React; simple ReAct loop built with ADK; single bash tool; no cyber-specific tools; general system prompt | 74.7 | |
| 23 | Grok 4.6 | high · Claude Code | 71.6 | |
| 24 | Sonnet 4.5 | high · Wiz Cyber Model Arena ADK React; simple ReAct loop built with ADK; single bash tool; no cyber-specific tools; general system prompt | 71.0 |
