Vals AI test finds open and closed models near parity on cyberattacks
What happened
On November 7, The Decoder reported results from Vals AI's CyberOffense test. AI agents broke into simulated corporate networks without human help, and open-weight models came close to restricted models, narrowing the capability gap. GPT-6 Sol reached 80.8% of its objectives and Deepseek V4.1 Flash 1017 reached 72.8%; a single run cost about $22 and $3 respectively.
Written by AI from the coverage · updated 9 hours ago
Coverage
Follow the reports to see the story from different sides.
- The DecoderThe gap between open and closed AI models in cyber capabilities keeps shrinking
Vals AI 的 CyberOffense 测试显示，AI 智能体能无人协助地入侵模拟企业网络，开放权重模型的攻击能力接近受限模型。GPT-6 Sol 达成 80.8% 的目标，Deepseek V4.1 Flash 1017 达成 72.8%，单次运行成本分别约为 $22 和 $3。
Heat over time
Heat now 8·Comparable peak 10（Nov 8 00:00）·Comparable change over 24 hours –
The trend only compares accounts observed without gaps, so its range may be smaller than the current heat. Hover or tap the chart for each hour; the left and right arrow keys step through it.