unbiased ai

Your Dify workflows, running on Pareto

You built the canvas node by node, and the classifier, the knowledge base and the prompts are where you want them. Keep all of it and change the model in each LLM node. Add Pareto through Dify’s OpenAI-API-compatible plugin and every node that uses it sends its request to Pareto, which selects frontier and open models and blends their outputs into one answer.

Choose Personal · $10/month See the setup →

Settings from Dify’s official plugin50M tokens a week

An illustration of Dify on Pareto. Not a screenshot.

Set it up with Dify’s OpenAI-API-compatible plugin

Since Dify 1.0, model providers are plugins. Install the official OpenAI-API-compatible plugin by langgenius, add Pareto as a model with your Unbiased key, then choose it in any LLM, Question Classifier or Agent node.

  1. Get an Unbiased API key

    Choose a plan, then create a key in the Unbiased platform. New accounts are reviewed and approved by a person before the key works.

  2. Install the plugin

    In Dify, open Integrations > Model Provider (on older self-hosted releases, Settings in the avatar menu > Model Provider) and install OpenAI-API-compatible (by langgenius) from the Marketplace. It does not come built in. Only workspace owners and admins can add model providers.

  3. Add Pareto as a model

    On the plugin’s card, choose Add Model. Set Model Type to LLM, type pareto as the Model Name, paste your key into API Key and set API Base URL to https://api.unbiased.ai/v1. Keep the /v1: the plugin adds /chat/completions itself.

  4. Raise the limits and turn on tool calls

    Dify starts Model context size and Upper bound for max tokens at 4096. Set them to 1000000 and 131072, and set Function Call Type to Tool Call so Agent nodes can use tools. Dify sends a test request when you save, so a wrong key or URL shows up right away.

  5. Pick it in your nodes

    In each LLM, Question Classifier or Agent node, open the model selector and choose Pareto (your display name) under OpenAI-API-compatible. Then click Preview and send a test message.

Settings follow Dify’s own documentation: Dify Marketplace: OpenAI-API-compatible plugin.

Add Model: OpenAI-API-compatible
Model Type: LLM
Model Name: pareto
Model display name: Pareto
API Key: your-unbiased-api-key
API Base URL: https://api.unbiased.ai/v1
Completion mode: Chat
Model context size: 1000000
Upper bound for max tokens: 131072
Function Call Type: Tool Call
API Type: Chat Completions API (/chat/completions)
# Leave the remaining fields at their defaults.

Leave Vision Support, Structured Output and Thinking Mode Support at their defaults unless you have checked the feature works for your app. Format compatibility is not a promise that every Dify feature works with Pareto.

The canvas stays. The model changes.

You add one model under one plugin and pick it in your nodes. The rest of the app is left as you built it.

One model string, several models behind it

Pareto is Unbiased’s blended AI model. On each request it selects frontier and open models and blends their outputs into one answer, so Dify talks to one model string, pareto, and never has to pick.

“When the frontier improves, Pareto improves; when frontier prices fall, you should benefit.” From the pricing page.

How a Pareto answer gets made →
An illustration, not a wiring diagram. The blend includes frontier and open models, and its composition can change.

What an agent task costs on Pareto

A chatflow can call the model more than once per message: once to classify, again to draft. The price of a finished task matters more than the price of a token, and the model card publishes both for Pareto on two public agent benchmarks.

Terminal-Bench 4.0Tasks an agent completes in a terminal
$0.48/task
50.8% solved
DeepSWE v1.1Changes an agent makes in a real codebase
$0.24/task
69.9% solved

Pareto 26.10, preliminary October 1, 2026 runs that may change; denominators and the cost method still require confirmation. Cost is the reported mean USD per task at API rates, not an estimate for your workload. Model card and method.

Pay for Dify by the month

One API key works across Dify and any other tool you point at Pareto. Pick a monthly allowance, or pay per token on Team.

Personal

$10/month

  • 50M tokens a week
  • 20 requests per minute
  • For one person
Choose Personal

Personal Max

$100/month

  • 600M tokens a week
  • 60 requests per minute
  • For one person
Choose Personal Max

Team

Pay as you go

  • Prepaid credits at published rates
  • 300 requests per minute
  • No weekly token cap
Get started with Team

Subscriptions are for one person and are not unlimited. New API accounts are reviewed by a person before access turns on, and commercial or production use needs a separate written agreement. Rates and limits are on Pricing.

Dify on Pareto, questions answered

Why do I have to type the model name?

Dify’s OpenAI-API-compatible plugin asks for the Model Name rather than fetching a list, and Unbiased’s published docs say model listing is not offered. Type pareto exactly, in lower case. The Model display name is only the label Dify shows in its menus.

Will Agent nodes use Pareto’s tool calls?

Only if you say so. The plugin sets Function Call Type to Not Support by default, which tells Dify the model can’t take tool definitions. Edit the model, set it to Tool Call and save. Pareto accepts tool definitions in the OpenAI Chat Completions format, but we have not run Dify against it, so tell us if a tool-using app misbehaves.

My long knowledge-base prompts get cut off. Why?

The plugin assumes a 4096-token context and a 4096-token output cap until you change them. Edit the model and set Model context size to 1000000 and Upper bound for max tokens to 131072, then set the max tokens you want in each LLM node.

Does this work on Dify Cloud and self-hosted Dify?

The steps are the same on both, since each installs model plugins from the Dify Marketplace. A self-hosted server needs to reach api.unbiased.ai over HTTPS. If your Dify app serves customers, that is commercial use, which needs a separate written agreement; the platform teams page covers it.

Can I switch back to another model?

Yes. Choose a different model in each node’s model selector. Pareto stays listed under the plugin until you remove it, so you can switch between the two and compare them in Preview.

Is Pareto made by the Dify team?

No. Pareto is Unbiased’s model. Dify is made by LangGenius, and Unbiased is not affiliated with or endorsed by them. You keep Dify and point it at the Unbiased API.

What happens when I hit the weekly allowance?

Requests over your plan’s limits return an HTTP 429 error. Personal includes 50M tokens a week and 20 requests per minute. Personal Max raises that to 600M a week and 60 requests per minute for $100/month. Team uses prepaid pay-as-you-go credits with no weekly token cap.

Do I choose between models?

No. You send every request to one model string, pareto. Pareto selects frontier and open models for each request and blends their outputs into one answer.

Is my code kept?

Personal and Personal Max are not our zero data retention tier: prompts and responses may be kept up to 30 days, as section 5(c) of the Terms explains. For work that must not be kept, ask about the Zero Data Retention API on the Private AI page.

Pareto works in other tools too

The same key and the same model string go into each of these. See them all.

Product names belong to their owners. Except where a guide says otherwise, Unbiased is not affiliated with or endorsed by the makers of the tools on this page.

Hand the triage queue to a new model

One plugin, one model name and the same canvas. Pareto starts at $10 a month for 50M tokens a week.

Choose Personal · $10/month See the setup →