According to CoinMeta, Qianwen’s three rounds of vulnerability testing have achieved a recall rate on par with Opus at 5%, and the cost is only half of the latter’s. The security company Aikido used code auditing Agent to test 7 models, including Qwen3.8-Max, Opus, Kimi K3 Max, DeepSeek V4 Flash, as well as GPT-5.6, SOL, Luna, and Terra. The tests covered 32 recently disclosed vulnerabilities, with each model running three times. In total, Qwen3.8-Max identified 26 vulnerabilities across the three rounds, achieving a recall rate of 81.3%, which is tied for first place with Opus. DeepSeek V4 Pro found 87.5% of the vulnerabilities in the three rounds of testing, surpassing Claude Opus and Qwen3.8 as well as Max. What’s more noteworthy is that DeepSeek costs an average of only $7.34 per vulnerability found, while Opus costs $66.97 per vulnerability, meaning the former’s cost is approximately 1/9 of the latter’s.