FrontierJuly 26, 2026via MarkTechPost
Sakana AI Releases Fugu-Cyber: An Orchestration Model Reporting 86.9% on CyberGym and 72.1% on CTI-REALM
Why it matters
Sakana AI is entering the specialized model wars with a security-focused orchestration model that benchmarks ahead of closed competitors. The gated access and defensive-use policy signal a new playbook for domain-specific model releases.
Key signals
- Fugu-Cyber scores 86.9% on CyberGym benchmark
- Fugu-Cyber scores 72.1% on CTI-REALM benchmark
- Outperforms GPT-5.5-Cyber and Claude Mythos Preview
- Access gated behind manual approval and defensive-use policy
- Sakana AI's Fugu orchestration model platform
- Security-tuned endpoint release
- Access restricted to Token Plan holders
The hook
86.9% on CyberGym. Sakana AI's Fugu-Cyber just outpaced GPT-5.5-Cyber on security benchmarks—and it's gated.
Sakana AI has released Fugu-Cyber, a security-tuned endpoint on its Fugu orchestration model. It reports 86.9% on CyberGym and 72.1% on CTI-REALM, edging past GPT-5.5-Cyber and Claude Mythos Preview. Access is gated behind manual approval, a defensive-use policy, and the Token Plan. Here is what the…