# CyberGym leaderboard

Agentic benchmark

> Every figure below is reproduced as its upstream source published it: nothing is modelled, estimated, interpolated or converted. `Unavailable` means no source published the value — it is never a zero. Each value carries its evidence state and the date it was observed.

Page: https://modelchoice.dev/leaderboards/benchmarks/agentic-cyber-gym

Every model the catalog carries a published CyberGym value for, ranked by that value.

## Benchmark record

| Field | Value |
| --- | --- |
| Category | Agentic |
| Measure | Vulnerability reproduction and PoC generation |
| Tasks | 1,507 vulnerability analysis instances |
| Difficulty | Real-world cybersecurity |
| Published by | [CyberGym: Evaluating AI Agents' Real-World Cybersecurity Capabilities at Scale](https://www.cybergym.io/) |

A cybersecurity task benchmark for evaluating defensive cyber workflows and vulnerability-oriented agent performance.

## CyberGym ranking

31 models with a published CyberGym value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.

| Rank | Model | Provider | Vulnerability reproduction and PoC generation | Evidence |
| --- | --- | --- | --- | --- |
| 1 | [MiMo-V2.6-Flash](https://modelchoice.dev/models/mimo-v2-6-flash) | Xiaomi | 95.1 | Observed 2026-11-21 · source benchlm:models |
| 2 | [MiMo-V2.6-Pro](https://modelchoice.dev/models/mimo-v2-6-pro) | Xiaomi | 94.0 | Observed 2026-11-21 · source benchlm:models |
| 3 | [DeepSeek V4.1 Flash 1017](https://modelchoice.dev/models/deepseek-v4-1-flash) | DeepSeek | 88.1 | Observed 2026-11-21 · source benchlm:models |
| 4 | [Ling 3.1 Flash](https://modelchoice.dev/models/ling-3-1-flash) | InclusionAI | 87.9 | Observed 2026-11-21 · source benchlm:models |
| 5 | [Fugu Cyber](https://modelchoice.dev/models/sakana-fugu-cyber) | Sakana AI | 86.9 | Observed 2026-11-21 · source benchlm:models |
| 6 | [Atria Dawn Preview](https://modelchoice.dev/models/atria-dawn-preview) | Shanghai Artificial Intelligence Laboratory | 86.5 | Observed 2026-11-21 · source benchlm:models |
| 7 | [Gemini 3.8 Flash Cyber](https://modelchoice.dev/models/gemini-3-8-flash-cyber) | Google | 86.2 | Observed 2026-11-21 · source benchlm:models |
| 8 | [Step 5 Preview](https://modelchoice.dev/models/step-5-preview) | StepFun | 84.7 | Observed 2026-11-21 · source benchlm:models |
| 9 | [GLM-5.3](https://modelchoice.dev/models/glm-5-3) | Z.AI | 84.5 | Observed 2026-11-21 · source benchlm:models |
| 9 | [GPT-5.6 Sol](https://modelchoice.dev/models/gpt-5-6-sol) | OpenAI | 84.5 | Observed 2026-11-21 · source benchlm:models |
| 11 | [Claude Mythos 5](https://modelchoice.dev/models/claude-mythos-5) | Anthropic | 83.8 | Observed 2026-11-21 · source benchlm:models |
| 12 | [DeepSeek V4 Pro](https://modelchoice.dev/models/deepseek-v4-pro) | DeepSeek | 83.3 | Observed 2026-11-21 · source benchlm:models |
| 12 | [DeepSeek V4 Pro 0813](https://modelchoice.dev/models/deepseek-v4-pro-max) | DeepSeek | 83.3 | Last good 2026-11-18 · source benchlm:models |
| 14 | [Gemini 3.5 Flash Cyber](https://modelchoice.dev/models/gemini-3-5-flash-cyber) | Google | 83.2 | Observed 2026-11-21 · source benchlm:models |
| 15 | [Claude Mythos Preview](https://modelchoice.dev/models/claude-mythos-preview) | Anthropic | 83.1 | Observed 2026-11-21 · source benchlm:models |
| 16 | [MiniMax M3.1 Flash](https://modelchoice.dev/models/minimax-m3-1-flash) | MiniMax | 82.2 | Observed 2026-11-21 · source benchlm:models |
| 17 | [GPT-5.5](https://modelchoice.dev/models/gpt-5-5) | OpenAI | 81.8 | Observed 2026-11-21 · source benchlm:models |
| 17 | [GPT-5.6 Terra](https://modelchoice.dev/models/gpt-5-6-terra) | OpenAI | 81.8 | Observed 2026-11-21 · source benchlm:models |
| 19 | [GPT-5.4](https://modelchoice.dev/models/gpt-5-4) | OpenAI | 79.0 | Observed 2026-11-21 · source benchlm:models |
| 20 | [Hy4 preview](https://modelchoice.dev/models/hy4-preview) | Tencent | 78.4 | Observed 2026-11-21 · source benchlm:models |
| 21 | [GPT-5.6 Luna](https://modelchoice.dev/models/gpt-5-6-luna) | OpenAI | 77.9 | Observed 2026-11-21 · source benchlm:models |
| 22 | [DeepSeek V4 Flash 0731](https://modelchoice.dev/models/deepseek-v4-flash-max) | DeepSeek | 76.7 | Observed 2026-11-21 · source benchlm:models |
| 23 | [Claude Opus 4.7](https://modelchoice.dev/models/claude-opus-4-7) | Anthropic | 73.1 | Observed 2026-11-21 · source benchlm:models |
| 23 | [Claude Opus 4.7 (Adaptive)](https://modelchoice.dev/models/claude-opus-4-7-max) | Anthropic | 73.1 | Last good 2026-11-18 · source benchlm:models |
| 25 | [GLM-5.1](https://modelchoice.dev/models/glm-5-1) | Z.AI | 68.7 | Observed 2026-11-21 · source benchlm:models |
| 26 | [Claude Opus 4.6](https://modelchoice.dev/models/claude-opus-4-6) | Anthropic | 66.6 | Observed 2026-11-21 · source benchlm:models |
| 27 | [Claude Sonnet 4.6](https://modelchoice.dev/models/claude-sonnet-4-6) | Anthropic | 65.2 | Observed 2026-11-21 · source benchlm:models |
| 28 | [Muse Spark 1.1](https://modelchoice.dev/models/muse-spark-1-1) | Meta | 59.0 | Observed 2026-11-21 · source benchlm:models |
| 29 | [Claude Opus 4.5](https://modelchoice.dev/models/claude-opus-4-5) | Anthropic | 50.6 | Observed 2026-11-21 · source benchlm:models |
| 30 | [Muse Spark](https://modelchoice.dev/models/muse-spark) | Meta | 43.5 | Observed 2026-11-21 · source benchlm:models |
| 31 | [GLM-5](https://modelchoice.dev/models/glm-5) | Z.AI | 43.2 | Observed 2026-11-21 · source benchlm:models |

Rows are ordered by the value the source published, highest first. The source does not state whether a higher value is the better result, so this document does not either: for a benchmark that measures a rate of failure, read the table from the bottom.
