Hinoki

Hinoki is a frontier cyber AI lab focused on making offensive cybersecurity capabilities more accessible to support better defense. Its flagship product, Hinoki Offensive One (HO1), is described as the first open-access cyber model post-trained for exploit development. Designed to work through complex tasks step by step within user-defined policies, HO1 can be used by individuals and teams through coding agents and development tools, including Claude Code, Codex, OpenCode, VS Code, and Cursor.

Hinoki presents HO1 as an accessible alternative to gated cyber models, highlighting its performance on the CyberGym benchmark. The company reports a score of 88.4%, which it says is 2.8 points higher than OpenAI’s dedicated GPT-5.5-Cyber model. The model is positioned for practical cybersecurity workflows, with the aim of helping users assess and address security challenges. Hinoki also offers an enterprise model, indicating services for organizations alongside access to its flagship model. Based in San Francisco, the company provides information about its products, partnerships, careers, and policies through its website.