We're excited to share that Qubrid AI is now officially listed on the Dify Marketplace. Dify developers can now add Qubrid as a model provider directly from the marketplace, with no custom integration code, no separate infrastructure, and no rewriting of existing prompts or agent logic.

Install the plugin, connect an API key, and Qubrid's model catalog becomes available across every chat app, agent, and workflow built in Dify.

Why This Listing Matters

Dify has become one of the most widely used platforms for building LLM-powered applications without having to hand-roll the orchestration layer every time. Teams use it to build internal copilots, customer-facing support agents, document and RAG pipelines, and multi-step automation workflows, all without writing the plumbing that connects a model to an app from scratch.

What Dify deliberately doesn't do is lock builders into one model provider. The platform is designed so the orchestration layer and the inference layer are separate decisions: pick the builder environment that fits your team, and pick the model provider based on cost, latency, model selection, and reliability, independently of each other.

That's exactly the gap Qubrid AI is built to fill. Qubrid operates as a GPU cloud and AI inference platform, offering OpenAI-compatible serverless model access alongside bare metal and on-premises options for teams that need more control. Because the API is OpenAI-compatible by default, plugging Qubrid into a platform like Dify doesn't require Dify to make any special accommodations, and it doesn't require Qubrid users to rewrite a single prompt or tool call. The interface is already shared.

That compatibility is what made this listing possible in the first place, and it's also what makes it useful day to day: any Dify builder who's used an OpenAI-style provider before already knows how to use Qubrid.

What's Included in the Plugin

The Qubrid AI plugin registers Qubrid as a model provider inside Dify, meaning it shows up anywhere Dify supports LLMs, including chat apps, agents, and workflows. Specifically, it includes:

• 43 predefined chat models, synced directly from Qubrid's live model catalog and available right in Dify's model picker, no manual configuration needed for the most commonly used models

• Full OpenAI API compatibility, so prompts, system messages, tool-calling configurations, and agent logic that were built for an OpenAI-style provider carry over unchanged

• Custom model support, so any Qubrid model ID not yet in the predefined list can still be added and used

• Serverless inference, meaning there's no GPU provisioning, capacity planning, or infrastructure management required on the user's end

• Day 0 access to new releases, including recent launches like GLM-5.2, where Qubrid has served as a launch partner

Connection requirements are minimal: an outbound HTTPS connection on port 443 to Qubrid's inference and catalog API. No inbound ports, no webhooks, no firewall changes beyond standard outbound access.

Setting It Up

Getting Qubrid running inside Dify takes about five minutes:

1. In Dify, go to Plugins → Install Plugin → Install from Marketplace, and search for Qubrid AI

2. Navigate to Settings → Model Provider → Qubrid AI → Set Up

3. Generate an API key from platform.qubrid.com/api-keys

4. Enter the key in Dify. Leave the API Base URL on its default unless you specifically need to override it

5. Save the configuration, then select any Qubrid model from the model picker in a chat app, agent, or workflow

From there, every app or workflow using that model provider routes through Qubrid's inference layer automatically.

Where This Fits for Builders

A few scenarios where this combination is especially useful:

Teams standardized on Dify, evaluating model providers. If you've already built workflows or agents in Dify and you're comparing inference providers on cost or model availability, switching to Qubrid is a model provider change, not a platform migration. Existing apps and workflows don't need to be rebuilt.

Teams that want early access to new model releases. Because Qubrid has launched as a Day 0 partner for models like GLM-5.2, Dify builders get access to new models as soon as they're available, rather than waiting for a separate integration cycle from another provider.

Teams running cost-sensitive or high-volume agent workloads. Serverless inference through Qubrid means there's no idle infrastructure cost between requests, which matters for agents that run intermittently or at unpredictable volumes.

Developers who want flexibility without lock-in. Because the API is OpenAI-compatible, nothing about adopting Qubrid inside Dify creates a new dependency that's harder to unwind than switching any other provider would be.

What's Next

This marketplace listing is a starting point, not a one-time integration. As Qubrid's model catalog grows and new releases ship, those models will be available to Dify users through the same plugin, without requiring an update on the Dify side. We'll continue to expand the predefined model list and keep the catalog in sync as new releases come online.

Get Started

The Qubrid AI plugin is live now on the Dify Marketplace. If you're building in Dify, adding Qubrid as a model provider takes a few minutes from start to finish. Grab an API key at platform.qubrid.com/api-keys and connect it to your next chat app, agent, or workflow.