OrcaRouter has released OrcaCyber Zero 1.5, a cybersecurity model built around a one-million-token context window.
Long context is the point of the release. Security work is document-heavy by nature, and analysts spend their time correlating logs, advisories, configuration files and incident notes that individually fit in a prompt but collectively never did.
A model that swallows a million tokens can hold a large slice of that material in one pass instead of stitching together retrieval results and hoping the right passage surfaces.
The tradeoff is familiar. Long windows degrade in practice, with attention thinning over distance, and a model that accepts a million tokens is not the same as one that reasons well across all of them.
The wider trend is clear either way. Labs are shipping domain-tuned security models rather than pointing general assistants at SIEM data, betting that specialized training beats scale on narrow detection work.
OrcaRouter has not published independent benchmark results, so the model’s accuracy against real detection workloads remains unverified.
