# Cyber Composite Index

> Vulnerability detection, security patching, and attack-defense CTF benchmarks.

- Canonical URL: https://prochoice.industries/cyber
- Domain JSON endpoint: https://prochoice.industries/api/v1/domains/cyber
- Epoch version: `cyber-v1.0.0`
- Snapshot hash: `43237d0f0a7497fc`
- Updated: 2026-11-21
- Cost axis: Measured cost per task across Cyber benchmarks (USD, log scale)
- Evaluated entries: 20

## Headline picks

- Budget: Gemini 3.7 Flash at 45.3 and $3.49
- Best value: GPT-6 Sol at 66.4 and $4.57
- Top score: Claude Fable 5.2 at 79.3 and $67.0

## Pareto frontier

| Rank | Model | Effort | Provider | Score | Cost / task |
| :--- | :--- | :--- | :--- | :--- | :--- |
| 1 | Claude Fable 5.2 | High | Anthropic | 79.3 | $67.0 |
| 2 | Gemini 4 Neon | High | Google | 73.6 | $55.7 |
| 3 | Gemini 4 Hydrogen | High | Google | 70.3 | $22.4 |
| 4 | Claude Opus 5.5 | High | Anthropic | 69.1 | $15.8 |
| 5 | GPT-5.6 Sol | Medium | OpenAI | 67.8 | $8.53 |
| 6 | GPT-6.2 Sol | High | OpenAI | 67.7 | $6.91 |
| 7 | GPT-6 Sol | Medium | OpenAI | 66.4 | $4.57 |
| 8 | GPT-6.1 Terra | High | OpenAI | 60.1 | $3.76 |
| 9 | Gemini 3.7 Flash | High | Google | 45.3 | $3.49 |

## Active benchmarks

- CWE-bench v2 (2.0, ACTIVE): https://cwe-bench.com/
- CyberArena Attack-Defense CTF (mc-v1, ACTIVE): https://cyber-arena.ai/
- Aikido Arena CVE Detection (1.0, ACTIVE): https://www.aikido.dev/aikido-arena

Full JSON dataset: https://prochoice.industries/api/v1/snapshot
