---
title: "CyberBench - Benchmark Evaluation - FinanceBenchmark"
description: "Evaluates cybersecurity defense, vulnerability discovery, and exploit mitigation."
canonical: "https://financebenchmark.ai/benchmarks/cyberbench"
type: "text/markdown"
---
# CyberBench

> Evaluates cybersecurity defense, vulnerability discovery, and exploit mitigation.

## Table of Contents
- [1. Specifications & Metadata](#specifications)
- [2. Model Rankings on this Benchmark](#rankings)
- [3. Source Attribution & Citation](#attribution)

## 1. Specifications & Metadata {#specifications}

- **Slug:** `cyberbench`
- **Domain Categories:** forensic
- **Composite Weight:** 10 (Tier 3)
- **Task Count:** Variable
- **Source / Publisher:** VALS (2026)

## 2. Model Rankings on this Benchmark {#rankings}

| Rank | Model | Organization | Overall | Eval Date | Source |
|:----:|:------|:-------------|:---:|:---------:|:------:|
| 1 | [GPT-5.6 Sol](https://financebenchmark.ai/models/gpt-5-6-sol.md) | OpenAI | **88.1** | 2026-07 | [Ref](https://vals.ai) |
| 2 | [GPT-5.6 Luna](https://financebenchmark.ai/models/gpt-5-6-luna.md) | OpenAI | **83.9** | 2026-07 | [Ref](https://vals.ai) |
| 3 | [Kimi K3](https://financebenchmark.ai/models/kimi-k3.md) | Moonshot AI | **79.0** | 2026-07 | [Ref](https://vals.ai) |
| 4 | [DeepSeek V4 Flash](https://financebenchmark.ai/models/deepseek-v4-flash.md) | DeepSeek | **77.5** | 2026-07 | [Ref](https://vals.ai) |
| 5 | [GLM 5.2](https://financebenchmark.ai/models/glm-5-2.md) | Zhipu AI | **77.4** | 2026-06 | [Ref](https://vals.ai) |
| 6 | [Gemini 3.5 Flash Lite](https://financebenchmark.ai/models/gemini-3-5-flash-lite.md) | Google | **72.3** | 2026-07 | [Ref](https://vals.ai) |
| 7 | [Inkling Small](https://financebenchmark.ai/models/inkling-small.md) | Thinking Machines | **68.8** | 2026-07 | [Ref](https://vals.ai) |
| 8 | [Inkling](https://financebenchmark.ai/models/inkling.md) | Thinking Machines | **64.1** | 2026-07 | [Ref](https://vals.ai) |
| 9 | [Gemini 3.6 Flash](https://financebenchmark.ai/models/gemini-3-6-flash.md) | Google | **48.5** | 2026-07 | [Ref](https://vals.ai) |
| 10 | [Claude Opus 5](https://financebenchmark.ai/models/claude-opus-5.md) | Anthropic | **40.7** | 2026-07 | [Ref](https://vals.ai) |
| 11 | [Qwen 3.7 Plus](https://financebenchmark.ai/models/qwen-3-7-plus.md) | Alibaba | **37.5** | 2026-06 | [Ref](https://vals.ai) |

## 3. Source Attribution & Citation {#attribution}

All scores for CyberBench are derived directly from published literature or reproducible evaluation harnesses.
To submit updated evaluations, email [submit@financebenchmark.ai](mailto:submit@financebenchmark.ai).