A+
95%+
Exceptional. Every dimension firing. Rare and earned.
A
90%+
Multi-agent gold. Genuine insight no single model could produce.
A-
85%+
Excellent. Strong divergence, thorough synthesis.
B+
80%+
Very good. Minor gaps in coverage or depth.
B
70%+
Solid consultation. Try mixing providers or increasing depth.
B-
60%+
Decent. Agents may be too similar or mediator too shallow.
C+
55%+
Below average. Check provider diversity and depth mode.
C
45%+
Agents too similar. Switch providers or add different roles.
C-
35%+
Poor quality. Likely same provider on all agents.
D
25%+
Not worth the cost. Rethink your agent configuration.
F
<25%
System failure or degenerate configuration.
How Agent Roles Improve Your Grade
Divergence -- Mix providers. Claude + GPT-4o + Gemini disagree at the training-data level. Same provider twice wastes the slot.
Depth -- Use Deep mode. Roles like Architect, Critic, and Patent Agent naturally produce structured reasoning with tradeoffs.
Coverage -- Spread across categories. Security + Cost Analyst + UX covers clusters that Dev + Dev + Dev never will.
Resolution -- Stronger agent disagreement gives the mediator more to resolve. Critic and Research roles provoke the best clashes.
Actionability -- Build and Full Stack missions produce phased plans with tasks. Brainstorm missions score lower here by design.