Jarvis Docs
Architecture

Cortex Mode Capability Matrix

Capability differences across Cortex tiers

Cortex Mode Capability Matrix

The Cortex tier determines orchestration strength, reflection depth, and policy strictness.

Each tier modifies behavior, not architecture.


Mode 0 — Off

  • No orchestration
  • No reflection
  • No memory gating
  • Direct model interaction
  • Tools disabled or rule-based only

Use case: direct chat mode


Mode 1 — Mobile

  • Single-pass routing
  • Minimal reflection
  • Lightweight memory triage
  • No multi-step planning
  • Limited retrieval evaluation

Target: 1–4B class model


Mode 2 — Low Spec PC

  • Reliable routing
  • Basic reflection pass
  • Controlled memory candidate generation
  • Retrieval trigger evaluation
  • Simple tool validation

Target: 5–8B class model


Mode 3 — Mid Spec PC

  • Two-pass reflection (draft + critique)
  • Better escalation detection
  • Structured memory validation
  • Improved policy enforcement
  • Tool result validation

Target: 10–14B class model


Mode 4 — High Spec PC

  • Multi-pass reflection
  • Deep planning capability
  • Strong conflict detection
  • Aggressive hallucination checking
  • Context optimization decisions

Target: 20B+ class model


Mode 5 — Remote Agent

  • Highest reasoning depth
  • Advanced reflection
  • Full orchestration strength
  • May operate with extended context
  • Dependent on remote provider

Privacy posture depends on configuration.


Important Constraint

Regardless of tier:

  • Only the Cortex may authorize tools
  • Only the Cortex may persist memory
  • Specialists remain advisory

Tiers increase strength, not authority.

On this page