AI agents can break into simulated corporate networks without human help, according to a new benchmark from Vals AI. The Vals CyberOffense benchmark places each agent alone on an isolated attack machine, starting with only a public web address.
GPT-6 Sol, part of OpenAI's restricted Daybreak cyber program, posted the best result at 80.8 percent of objectives. Deepseek V4.1 Flash 1017, a freely available open-weight model, scored 72.8 percent. The report says it costs about $3 per run, compared with roughly $22 for GPT-6 Sol. Vals researchers say that once a model's weights are public, no provider can control or restrict how it is used.
In the harder Lighthouse test, Daybreak took admin control of the first company, stole passwords, and used them to break into the second company. It completed 27 of 31 objectives in under three hours. Deepseek completed 21 objectives at a tenth of the cost. The report says it was not limited solely by safety guardrails.
Runtime also affected results. Deepseek scored zero in its first 25 million tokens and made most of its progress only after 100 million tokens. According to Vals, short tests therefore underestimate what open models can do.
Vals says