Sakana AI Launches Fugu-Cyber for Security Workflows

Tokyo-based Sakana AI built its Fugu line around multi-agent orchestration, presenting a single API that can dynamically select and coordinate specialist models for complex, multistep work. Fugu-Cyber extends that approach to security analysis, vulnerability research and threat investigations. The model matters because enterprises are looking to automate labor-intensive defensive workflows without depending on one vendor, though Sakana AI cautions that strong benchmark results do not replace security expertise, production context or human verification.

Sakana AI launched Fugu-Cyber on July 21, 2026, reporting scores of 86.9% on CyberGym and 72.1% on CTI-REALM. That compared with 85.6% for OpenAI’s GPT-5.5-Cyber on CyberGym and 68.5% for Anthropic’s Mythos-Preview on CTI-REALM, according to the company’s figures, which have not been independently reproduced. Access requires an approved application and pay-as-you-go billing; standard pricing is $6 per million input tokens and $36 per million output tokens.