Skip to content

Vals CyberOffense benchmark: open-source models close in on closed-source cyberattack capability

What happened

AI digest

Vals AI has launched CyberOffense, a benchmark in which AI agents independently attack a simulated corporate network on an isolated Kali Linux machine, measuring their cyberattack capability by goal completion rate. Reporting on 7 November 2026 shows that GPT-6 Sol from OpenAI's restricted-network project Daybreak achieved an 80.8% goal completion rate, with the open-weight model Deepseek V4.1 Flash 1017 close behind at 72.8%, and the gap between them continuing to narrow. On cost, an open-source model run costs about $3, versus about $22 for a closed-source model. The benchmark results currently show that the gap in cyberattack capability between open-source and closed-source models is steadily shrinking.

Written by AI from the reports · updated 8 d ago

Timeline

Follow the reports to see the event from each side.

Nov 7
  1. The DecoderSelected
    The gap between open and closed AI models in cyber capabilities keeps shrinking

    In the Vals AI benchmark, OpenAI's GPT-6 Sol Daybreak hits 80.8 percent of attack objectives in simulated corporate networks. Worse, the open-weight model DeepSeek V4.1 Flash reaches 72.8 percent at a tenth of the cost. Because open models can no longer be restricted, analysts say the window for defenders is shrinking fast.

Heat of this event

Current heat 3·Peak in the comparable range 4(Nov 9 09:00)·Change over 24 hours in the comparable range –

01234Nov 909:00Nov 914:00Nov 918:00Nov 923:00

The trend compares only the same accounts observed without a break, so its range can be narrower than the current heat count. Move the pointer or tap the chart to see each hour; with a keyboard, use the left and right arrow keys.