FlashLabs Brings OrcaCyber Zero 1.5 to OrcaRouter
A security-focused model scoring near the top of vulnerability benchmarks, offered on a screening basis, moves AI-assisted vulnerability discovery from explanation toward autonomous testing, which is the part of the stack enterprises have been slowest to adopt.
Reporting from 1 source: ASCII.jp.
FlashLabs has begun offering OrcaCyber Zero 1.5, a cybersecurity-specialized frontier model, on OrcaRouter, the AI inference gateway developed by Continuum AI and provided by FlashLabs for the Japanese market. According to the developer, the model scored 95.8% on CVE-Bench and 100% on Cybench under unconstrained agent execution, and it supports a 1 million token context window.
The model succeeds OrcaCyber Zero 1.0 and keeps the previous version's OpenAI-compatible API, native tool calling, and structured JSON output. FlashLabs describes the intended work as investigating large codebases and complex attack surfaces while holding long contexts.
It is provided on a screening basis to trusted security researchers, red teams, and users doing authorized security testing. FlashLabs says the autonomous harness running the previous model ranks in the top 3 on CyberGym Level 1, a global benchmark that reproduces real-world vulnerabilities, per its own research as of October 11, 2026.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.