CyberBench v1.1 PoC
Agents
The 60-task proof-of-concept track of CyberBench v1.1, evaluated in offline sandboxes with a hardened clean-room grader. A submitted input must crash the vulnerable build and, when the differential oracle is valid, avoid crashing the fixed build.
Top models (higher is better)
| Model | Score |
|---|---|
| Muse Wonder | 75.0 |
| Gemini 4 Argon | 70.0 |
| Kimi K3 | 68.3 |
| GPT-6 Sol | 66.7 |
| GPT-6 Luna | 65.0 |
| MiMo-V2.6-Flash | 65.0 |
| GPT-5.6 Luna | 63.3 |
| GPT-5.6 Sol | 63.3 |
| Muse Spark 1.3 | 63.3 |
| Muse Wonder Full | 63.3 |
| Claude Haiku 5.5 | 61.7 |
| DeepSeek V4.1 Flash | 61.7 |
| Grok 4.5 | 61.7 |
| DeepSeek V4 Pro 0813 | 60.0 |
| MiMo-V2.6-Pro | 60.0 |
| DeepSeek-V4-Flash-0731 | 56.7 |
| GLM-5.3 | 56.7 |
| GPT-5.6 Terra | 56.7 |
| Grok 4.6 | 55.0 |
| Grok 4.7 | 55.0 |
| Inkling | 55.0 |
| Inkling-Small | 55.0 |
| MiniMax M3 | 55.0 |
| Ember-1 | 53.3 |
| Muse Spark 1.2 | 53.3 |
| Qwen3.8 27B | 51.7 |
| GLM-5.3 Flash | 50.0 |
| Gemini 3.5 Flash-Lite | 45.0 |
| Hy4 Preview | 45.0 |
| Sonnet 5 | 41.7 |
| Mercury 2.5 | 26.7 |
| Nemotron 3.5 Lightning 30B A3B BF16 | 25.0 |
| Gemini 3.6 Flash | 5.0 |
| Gemini 3.7 Flash | 0.0 |
| Gemini 3.8 Flash | 0.0 |
| GPT-6.1 Sol | 0.0 |
| GPT-6 Astra | 0.0 |