CyberGym — Cybersecurity vulnerability-reproduction agentic benchmark
| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | GLM-5.3 | Z.ai | 84.5 | 2 | 2026-08-23 |
| 2 | Claude Mythos 5 | Anthropic | 83.8 | 2 | 2026-08-23 |
| 3 | GPT-5.6 Sol | OpenAI | 83.6 | 1 | 2026-08-17 |
| 4 | Gemini 3.5 Flash Cyber | 83.2 | 1 | 2026-08-23 | |
| 5 | Claude Opus 4.8 | Anthropic | 83.1 | 4 | 2026-08-23 |
| 6 | GPT-5.5 | OpenAI | 81.8 | 2 | 2026-08-23 |
| 7 | Kimi K3 | Moonshot | 80 | 1 | 2026-08-13 |
| 8 | DeepSeek-V4-Flash | DeepSeek | 76.7 | 6 | 2026-08-23 |
| 9 | Claude Opus 4.6 | Anthropic | 73.8 | 4 | 2026-08-23 |
| 10 | Claude Opus 4.7 | Anthropic | 73.8 | 2 | 2026-08-23 |
| 11 | GLM-5.1 | Z.ai | 68.7 | 3 | 2026-08-23 |
| 12 | GPT-5.4 | OpenAI | 66.3 | 2 | 2026-05-18 |
| 13 | Muse Spark 1.1 | Meta | 59 | 1 | 2026-07-13 |
| 14 | DeepSeek-V4-Pro | DeepSeek | 52.7 | 2 | 2026-08-13 |
| 15 | Claude Opus 4.5 | Anthropic | 50.6 | 3 | 2026-07-29 |
| 16 | GLM-5 | Z.ai | 48.3 | 3 | 2026-05-18 |
| 17 | Muse Spark | Meta | 43.5 | 1 | 2026-07-16 |
| 18 | Kimi K2.5 | Moonshot | 41.3 | 5 | 2026-08-23 |
| 19 | Gemini 3 Pro | 39.9 | 2 | 2026-08-24 | |
| 20 | Gemini 3.1 Pro | 38.8 | 2 | 2026-05-18 | |
| 21 | GLM-4.7 | Z.ai | 23.5 | 1 | 2026-06-13 |
| 22 | DeepSeek-V3.2 | DeepSeek | 17.3 | 4 | 2026-06-15 |
Best tracked score per model (default configuration; source-attributed and verification-tiered). Open a model for its full benchmark surface, provenance and pricing.