Sakana AI Releases Fugu-Cyber: Orchestration Model Scores 86.9% on CyberGym and 72.1% on CTI-REALM
By Asif Razzaq | July 25, 2026
Sakana AI has unveiled Fugu-Cyber, an advanced orchestration model designed for cybersecurity operations. The model achieved notable benchmark results: an 86.9% score on the CyberGym evaluation and 72.1% on the CTI-REALM dataset. This release comes amid a growing need for intelligent, automated security solutions in 2026, as organizations face increasingly sophisticated cyber threats.
Fugu-Cyber functions as an orchestration model, coordinating multiple AI agents to perform complex security tasks. This approach reflects the broader trend toward agentic AI systems that can autonomously handle threat detection, analysis, and response. By integrating large language models with specialized cybersecurity tools, Fugu-Cyber aims to reduce the burden on human analysts and accelerate incident response times.
The model's performance on CyberGym and CTI-REALM benchmarks underscores its capability in real-world cybersecurity scenarios. CyberGym evaluates proficiency in simulated cyber defense exercises, while CTI-REALM tests threat intelligence and reasoning skills. These scores position Fugu-Cyber as a competitive option for enterprises seeking AI-driven security orchestration.
Sakana AI's latest release highlights the convergence of agentic AI and cybersecurity, a rapidly expanding field in 2026. As organizations continue to adopt AI-powered defenses, models like Fugu-Cyber represent a significant step toward fully autonomous security operations centers.
via MarkTechPost