Continuum AI launches OrcaCyber Zero 1.5, a cutting-edge cybersecurity reasoning model.
The updated model achieves a score of 95.8% on CVE-Bench and 100% on CyBench, aiming to advance autonomous vulnerability discovery, exploit validation, and remediation.
San Francisco, October 12 /PRNewswire/ — Continuum AI today announced the launch of OrcaCyber Zero 1.5, its latest cutting-edge cybersecurity reasoning model designed to identify critical vulnerabilities, investigate multi-stage attack paths, and validate exploitability in real-world environments. Featuring a 1 million token context window, native tool calling, and enhanced reasoning capabilities, the model is built for complex autonomous security investigations within large-scale codebases and interconnected systems. Security teams can apply its reasoning power to intricate code repository structures, massive volumes of security logs, and extensive investigation histories without needing to split evidence across isolated sessions.
Benchmark performance
According to the company’s published evaluations, OrcaCyber Zero 1.5 achieves 95.8% on CVE-Bench (23 out of 24 evaluable tasks), 100% on Cybench (39 challenges, unconstrained), and 93.9% on HumanEval+. Its predecessor, OrcaCyber Zero 1.0, supported an autonomous framework that achieved over 98% pass@1 performance on CyberGym and, according to Continuum AI, ranked among the top three globally. Continuum AI also states that Zero 1.0 has provided security workload support for major software companies, government agencies, blockchain networks, and cybersecurity firms.
Six Controlled Security Demonstrations
- Admin takeover: Bypass authentication using SQL injection to gain admin access and expose user records.
- Sensitive file access: Locate restricted files and demonstrate access control bypass.
- Cross-site scripting: Verify JavaScript execution in a real browser.
- Business logic exploitation: Manipulating checkout quantities to create negative-value orders.
- AI-to-AI prompt injection: Extracting hidden instructions from another AI assistant in controlled tests.
- Self-repair: Detect vulnerabilities, reproduce them, generate patches, and verify the fixes through testing.
From Alerts to Actionable Network Defense
OrcaCyber is designed to prioritize reproducible evidence over alert volume. By integrating vulnerability discovery, exploit validation, root cause analysis, and tested remediations, it helps defenders focus on truly high-impact weaknesses while reducing false positives. Its broader goal is to advance security operations from passive scanning to autonomous investigation and defense, while maintaining human oversight for critical actions. This approach connects individual findings into coherent attack narratives, helping analysts distinguish theoretical vulnerabilities from verified risks and prioritize remediations supported by reproducible tests.
Continuum AI also plans to release next week a more powerful open-source cybersecurity model with support for on-premises deployment. The company states that this release aims to make cutting-edge cybersecurity intelligence more accessible while enabling organizations to maintain direct control over sensitive security workloads and infrastructure.
Availability and Company Information
Security researchers and enterprise teams can experience OrcaCyber Zero 1.5 on OrcaRouter and apply for model access. Continuum AI develops cutting-edge cybersecurity models and secure AI infrastructure, including the privacy-focused OrcaRouter AI gateway. For more information, visit continuum01.ai.
Media contact: Chris Zhang
650.609.7501
[email protected]