Classification System

Five classes. Every context.

AICONSORTIUM's models are organised into five capability classes, each engineered for a distinct purpose, from flagship multimodal intelligence to efficient edge deployment. Every model we ship is classified to help you choose the right capability for your context.

AICONSORTIUM classes visual

Capability tiers

Every scale, one architecture

From the frontier Oracle class to edge-deployable Echo, each tier sets a capability ceiling, compute footprint, and deployment context, so you run exactly the intelligence you need.

Request access
Oracle Class Clark

The flagship tier

Clark is Oracle Class, the highest capability tier in the AIC model system. Multimodal at its foundation and trained for maximum accuracy on complex tasks, Oracle-class models scale from a 70B backbone to 1T parameters. Built for tasks where error is costly and context is everything.

Multimodal foundation 70B to 1T scale Maximum accuracy Complex multi-step tasks
Explore Clark
Atlas Class Sattam.ai

The domain-expert tier

Atlas-class models are trained for deep expertise in specific verticals. Narrower than Oracle but deeper in their domain, and built for tasks where mastery matters more than breadth. Sattam.ai, our jurisdiction-aware legal intelligence model, is the first Atlas-class deployment.

Vertical specialisation Domain-native reasoning Deep over broad Expert-level accuracy
Explore Sattam.ai
Sage Class Sattam.ai

The reasoning tier

Sage-class models are built for extended chain-of-thought and deliberate reasoning. Slower and more methodical than Oracle, they excel at complex multi-step problems where the path to the answer is as important as the answer itself. Sattam.ai operates in this tier.

Extended chain-of-thought Deliberate reasoning Multi-step problems Methodical & thorough
Cipher Class Roadmap

The efficient tier

Cipher-class models are compressed and quantized for fast inference at the edge. Lower parameter count and high throughput make them deployable on constrained hardware, which suits offline environments, secure facilities, and field operations where connectivity cannot be assumed.

Compressed & quantized High throughput Edge deployable Offline capable
Register Interest
Echo Class Roadmap

The fine-tuned tier

Echo-class models are specialised through fine-tuning on domain-specific datasets. The fastest class to deploy for narrow tasks, for when you know exactly what you need and need it immediately. They deliver targeted performance without the compute overhead of a general-purpose model.

Fine-tuned on domain data Fastest to deploy Narrow task performance Low overhead
Register Interest

Not sure which class fits?

We'll help you choose

Our team will walk through your deployment requirements, sovereignty constraints, and capability needs to identify the right class and configuration for your use case.

Talk to our team View Current Lineup