> ## Documentation Index > Fetch the complete documentation index at: https://docs.trynito.ai/llms.txt > Use this file to discover all available pages before exploring further. # Fusion > One prompt to two or three models, plus a synthesis. Fusion sends one prompt to two or three models at once and returns each answer with a synthesized conclusion above them. It is part of every paid plan. ## Setting Up Open the model picker and use the **Fusion** card. Pick up to three models; rows that do not match the draft's privacy mode or the chat's protected history read **Incompatible**. **Apply Fusion** arms it, the composer shows **Fusion · N models**, and you send as normal. The result shows the synthesis on top and one card per model with an output preview, token count, latency, and any Web Citations. **Open full result** shows everything side by side. Copy any output on its own, or **Continue from** a model's output to fork the chat with that answer. Fusion stays armed until you turn it off or start a new chat. Each participant and the synthesis are billed as separate calls. ## Web Tools The chat's Web Search and Web Fetch settings apply to every participant; only tools every selected model supports are forwarded. Sources the participants return also reach the synthesis. ## Confidential Fusion A Fusion of Confidential models runs entirely on Confidential routes, and one participant acts as the synthesis model so the synthesis stays inside sealed hardware. Each card carries its own attestation, and a consolidated report follows. Confidential models cannot mix with Anonymous or Private ones; Anonymous and Private can mix. ## Encrypted Fusion Each participant request is encrypted in your browser and decrypted locally, so the synthesis cannot run on the server. The cards appear first with a **Synthesize in Encrypted** action. Clicking it fetches a fresh attestation for the synthesis model, sends the prompt plus the candidate answers through the Encrypted path, and replaces the summary with the decrypted synthesis. ## Retrying a Participant If one model fails, the other answers and the synthesis still render and the failed card shows a short reason. In the same session, **Retry** reruns that participant with web tools off. Turns with attachments, reopened saved chats, and shared snapshots regenerate the whole reply instead. ## Protected History and Plans A chat holding Confidential or Encrypted history accepts only a Fusion whose models meet that level. Fusion requires an active paid plan; if a plan lapses, past results stay readable while regenerate, synthesis, and retry wait for the plan. ## Next Where the Fusion card lives. The Fusion attestation report. Why Confidential stays isolated. Fusion from Plus. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.