You are selecting models for a multi-agent system: a lightweight cl…

Question

You are selecting models for a multi-agent system: a lightweight classifier agent routes intents; a deep reasoning agent handles complex multi-step plans; a vision agent reads PDF screenshots. What architectural principle should guide model assignment?

Accepted Answer

Match task demands to model family capacities: small/fast models for routing, stronger reasoning models for planning, and vision-capable models only where image understanding is required. — Correct answer: B. Match tasks to model families: small/fast for routing, strong reasoners for planning, vision only when needed. One giant model everywhere (A) wastes money and latency. Micro-models only (C) may destroy planning quality. Random rotation (D) is chaos. Document model SLOs per agent and fallback models when primary capacity is constrained.

Answer

Assign the single largest multimodal model to every agent regardless of task cost or latency.

Answer

Use only on-device micro-models for all planning to avoid any cloud dependency even if accuracy collapses.

Answer

Randomly rotate models per request without regard to capability to 'balance load.'

You are selecting models for a multi-agent system: a lightweight classifier agent routes intents; a deep reasoning agent handles complex multi-step plans; a vision agent reads PDF screenshots. What architectural principle should guide model assignment?

Short Explanation

Full Explanation