Skip to main content
Model Fusion sends one prompt to two or three models at once, then a judge model reads every answer and writes one merged reply. You get the synthesis and each participant’s output.

Where

Fusion needs a paid plan. See Plans.

How It Works

  1. Fan-out. Anonymous and Private models run concurrently; all-Confidential sets run one at a time inside attested routes.
  2. Synthesis. A judge model merges the answers. For all-Confidential sets, one Confidential participant is the judge, so the synthesis stays attested.
The result arrives once every participant and the synthesis finish. A failed participant is reported in its own entry.

Privacy Rules

  • Anonymous and Private models can mix.
  • Confidential models run only with other Confidential models, judge included.
  • Encrypted Fusion runs in the Browser App, with the synthesis as an explicit step.

From Your Own Code

Call POST /api/v1/chat/fusion with models and messages. The response includes each participant and the synthesis. Each participant and the synthesis meter separately. A paid plan with Fusion access is required; prepaid balance alone does not unlock it.

Fusion in the Browser App

Pick and compare.

Fusion in the Plugin

Claude Code and Codex.

Fusion API

HTTP examples and results.