Other models refuse offensive security tasks by design. alias models execute them — trained from the ground up for penetration testers, red teams, vulnerability researchers and security operations centers.
Two flagship models: alias2, our most capable model — reserved for government and critical infrastructure. alias2-mini, the best-performing cybersecurity model on workstations — available to every PRO subscriber and deployable on-prem. Currently #1 on CAIBench across all commercial LLMs.
On-prem · best on workstations
The model behind the #1 CAIBench result — and the best cybersecurity model that runs on a workstation. Available to every PRO subscriber as part of CSI and deployable on-premise for full sovereignty.
#1 Base CTF Rank — 71.3% pass@1Flagship cybersecurity model
Our most capable model, available to government and critical infrastructure operators only. Designed for hands-on red team operations, advanced exploit development, and adversary simulation under strict licensing.
Other models refuse.
alias models execute.
Every alias release is measured on the same security benchmarks and shipped through both cloud and on-premise routes — same weights, same capabilities, your choice of sovereignty.
alias.family
ALIAS_TRACK ∈ {flagship, mini}
127.0.0.1:PORTEvery alias release is measured on the same 33-challenge CAIBench-Jeopardy (Cybench) benchmark — pass@3, max 300 agentic interactions, 245 minutes and ≤$40 API per challenge.
Each row is an alias model. Each column is one Cybench challenge, ordered by difficulty (Beginner → Very Hard). A filled cell means the challenge was solved.
Source: Mayoral-Vilches et al. (2026). Towards Cybersecurity Superintelligence. arXiv:2601.14614
CAIBench-Jeopardy (Cybench) solve rate by launch date — pass@3, ≤300 agentic interactions, ≤$40 API per challenge.
Source: Mayoral-Vilches et al. (2026). Towards Cybersecurity Superintelligence. arXiv:2601.14614
Unlike US-based providers, alias models are trained and hosted exclusively in European infrastructure. Air-gapped on-premise available for organizations that require zero external dependencies.
Full data protection with audit trails
Enhanced incident reporting capabilities
On-premise for maximum sovereignty
Built for regulatory compliance
Model updates, competition results, and deployment milestones.
Get unlimited tokens via PRO access, or deploy alias on-premise for full data sovereignty. Either way, you get the only LLM family that won't refuse the job.