Vals tests put open-weight DeepSeek close to gated cyber model

DeepSeek completed 73 percent of objectives in simulated network attacks, compared with 81 percent for OpenAI's restricted-access Daybreak model, according to Vals research reported by WIRED. The tests covered two scenarios, including compromising one company network to reach another. Vals also reported token costs of $2 to $3 for DeepSeek attacks, compared with $17 to $46 for GPT-6-Sol. That cost comparison involves a different OpenAI model from the capability comparison. The results concern simulated networks; observed attackers often use older, less capable models.