Fusion

Multiple Models Collaborate, One Answer

Create custom fusion presets, call multiple AI models in parallel. Judge model intelligently fuses optimal results. One request, multi-model intelligence.

Create Your Fusion Preset in 3 Steps

1

Create a preset in Console 'Fusion' page

Select 2-5 models, specify a Judge Model

2

Use {ID Code}/{Preset Name} to call

Model Name
model: abc123/x4k7m
3

System calls models in parallel, Judge Model fuses the best answer

Combines strengths of multiple models for higher quality answers

How It Works

Parallel requests → Individual responses → Judge fusion → Optimal result

01

Parallel Calls

On receiving request, call all preset models simultaneously (max 3 concurrent), get independent responses from each

02

Collect Responses

Aggregate all model responses, preserving each model's unique strengths and perspectives

03

Judge Fusion

Judge Model reviews all responses, extracts optimal content, generates one high-quality fused answer

04

Return Result

Return the fused answer to the user, supports both streaming and non-streaming modes

Use Cases

When should you use Fusion?

Complex Reasoning

Math derivation, logical analysis and other difficult tasks. Let different models each try, Judge selects optimal solution. Higher success rate than single model.

Multi-Perspective Answers

Creative writing, brainstorming and other open tasks. Different models have different perspectives, fused content is more comprehensive and creative.

Reduce Error Rate

Critical decision scenarios. Multi-model cross-validation, Judge filters out errors, significantly reduces single model error risk.

High-Quality Translation

Each model translates once, Judge combines advantages of each version. Produces more faithful and elegant translation than any single model.

Parallel & Efficient

Models work simultaneously, no sequential waiting

Smart Judge

AI auto-selects the best, no manual intervention

Flexible Combinations

Freely combine 2-5 models

Protocol Compatible

OpenAI / Anthropic both callable

Billing

Each Model Billed Independently

Each participating model is billed at its own route rate. The Judge Model is also billed separately. Plan users consume plan calls.

Presets are Private

Each user's Fusion presets are completely private. Other users cannot access or call them.

FAQ

Q

How many models can I include?

Each Fusion preset supports up to 5 models. The system auto-limits concurrency to 3 (configurable via admin setting fusion_max_parallel). Models exceeding the concurrency limit are automatically trimmed.

Q

What does the Judge Model do?

The Judge model reviews all participating models' responses and synthesizes optimal content. We recommend choosing a strong reasoning model as judge (e.g. GLM-5.2, Kimi K2).

Q

Does it support streaming output?

Yes. Fusion internally calls models in parallel (forced non-streaming), then outputs the fused result via standard SSE streaming format. Cherry Studio, Cursor and other clients can receive it normally.

Q

Can other users access my Fusion presets?

No. Fusion presets are completely private. Only the owner can call them via {Your ID Code}/{Preset Name}.

Q

What is the ID Code in the call name?

The ID Code is your account's unique identifier (User.InviteCode). You can copy the full call name from the Console 'Fusion' page.

Create your first fusion preset

Go to Console 'Fusion' page and start collaborating with multiple models

Create Now