Fusion

When to use Fusion

Fusion runs multi-model deliberation: a panel answers in parallel, then a judge synthesizes consensus, contradictions, and blind spots into one answer. Reach for it on high-stakes questions — architecture decisions, tricky analysis, final review passes — where one model's blind spot is expensive. For everyday traffic, plain tier routing is faster and cheaper.

Requesting Fusion

Send the fusion alias like any other request:

{ "model": "intelliverse/fusion", "messages": [...] }

Cost

Fusion bills as a single credit-metered request (panel + judge included) at a flat rate per ~1k tokens:

• Fusion (advanced panel) — 20 credits per request (~1k tokens).
• Fusion (frontier panel) — 120 credits per request (~1k tokens).

See Credits & billing for how credits work, and /fusion for the product deep-dive.