Models

Cracken gives every operation a security-focused model portfolio: Cracken's own Red model plus curated frontier models organized as Max, Smart, and Efficient. Choose the capability you want when you launch or steer work, then adjust reasoning effort separately where the selected model supports it.

At a glance

• Red is built for security work — Cracken's own model with no cybersecurity refusal for authorized offensive-security operations.
• Capability tiers stay understandable — Max, Smart, and Efficient describe the job each model is meant to do instead of exposing a changing vendor catalogue.
• Smart is the recommended starting point — balanced reasoning makes it a strong fit for most proactive-security work.
• Alternatives and effort are separate controls — Alt selects an alternative model family; effort controls how much reasoning the selected model applies.
• Availability and relative cost are visible — the authenticated selector shows what your deployment can use before you start or steer an operation.

Choose a model

Model	What it gives you
Red	Cracken's own security-specialized model with no cybersecurity refusal for authorized offensive-security work. Use it when general-purpose models would block legitimate testing.
Max	Cracken's highest-capability tier for complex, ambiguous, or high-stakes security work where depth matters most.
Smart	Balanced reasoning for broad security operations. This is the recommended starting point for most work.
Efficient	Fast, lightweight execution for focused tasks, parallel work, and high-volume workflows.

This portfolio lets teams choose for the security task rather than learn a vendor's model catalogue. Cracken presents the curated models under consistent operational names even as the portfolio evolves.

Try an alternative model

Where an Alt option appears, it provides an alternative model family in the same capability category. Use it when you want a different approach to the same job without moving from, for example, Smart to Max.

Alt is not a reasoning-effort setting. Where effort is available, its separate selector lets you trade faster responses for deeper reasoning while keeping the same model selected.

Review relative cost

The model selector shows the current relative cost tier before you start or steer an operation. Use that indicator to balance capability and scale; actual credit use varies with the task, operation length, selected effort, and model behavior. On commercial deployments, review Settings → Billing for account usage and limits. Self-hosted deployments do not expose that tab; model usage and capacity are governed by the configured model infrastructure.

Next steps

• Operations — choose a model when you start or steer an operation.
• Billing — on commercial deployments, review your plan, credit usage, and available top-ups.