Skip to main content
Each Dify Cloud plan includes a set number of AI credits for using supported models hosted by Dify. They give you immediate access to models from different providers, so you can start building and testing apps without first adding your own API keys. App runs that use AI credits draw from the workspace’s shared balance, including runs from the editor, published web apps, the Service API, and triggers.

​
Understand AI Credit Usage

Your workspace may use AI credits whenever a supported model runs. Common cases include:
  • Running an app with an LLM, whether configured directly in a Chatbot, Text Generator, or Agent app or through an LLM, Question Classifier, Parameter Extractor, or Agent node in a Workflow or Chatflow.
  • Using AI-assisted generation in the app editor to create or revise prompts, code, structured output configurations, or Workflow and Chatflow drafts.
  • Indexing or searching a knowledge base, which may run an embedding model for indexing or vector search, an LLM to generate Q&A chunks from source documents, or a rerank model when reranking is enabled.
  • Transcribing voice input or generating audio output in an app with a speech-to-text or text-to-speech model.
Each model has a fixed credit cost that applies every time it runs, regardless of input or output length. Iterations, Loops, Agent reasoning, and knowledge base indexing may run a model more than once. For example, a Workflow uses a 5-credit LLM in an Iteration to analyze three customer reviews, then uses the same LLM to summarize the findings. The four runs use 20 AI credits in total.

​
Compare Model Costs

The following costs apply each time a model runs:
  • 25 credits: gpt-5.6-sol, gpt-5.5.
  • 20 credits: gpt-4.
  • 15 credits: gpt-5.6-terra, gpt-5.4.
  • 10 credits: gpt-5.2, gpt-4o, gpt-4o-transcribe-diarize, gpt-4o-transcribe, whisper-1, gpt-4o-mini-tts-2025-12-15, tts-1-hd, tts-1.
  • 5 credits: gpt-5.6-luna, gpt-5.4-mini, gpt-5, gpt-5.1, gpt-4.1, o3-mini, o4-mini, gpt-4o-mini-transcribe-2025-12-15, gpt-4o-mini-transcribe, text-embedding-3-small, text-embedding-3-large, text-embedding-ada-002.
  • 1 credit: gpt-5-mini, gpt-4.1-mini, gpt-4o-mini, gpt-3.5-turbo.
  • 20 credits: Claude Opus 4.7, Claude Opus 4.6, Claude Opus 4.5.
  • 10 credits: Claude Sonnet 4.6, Claude Sonnet 4.5.
  • 5 credits: Claude Haiku 4.5.
  • 10 credits: Gemini 3.5 Flash, Gemini 3.1 Pro-Preview.
  • 5 credits: gemini-3.7-flash, gemini-embedding-001, gemini-embedding-2.
  • 3 credits: Gemini 3 Flash-Preview.
  • 1 credit: Gemini 3.5 Flash-Lite, Gemini 3.1 Flash-Lite-Preview, Gemini 3.1 Flash-Lite, Gemini 2.5 Flash, Gemini 2.0 Flash, Gemini 2.0 Flash-Lite.
  • 20 credits: grok-4.5, grok-4.6.
  • 10 credits: grok-4.3.
  • 20 credits: qwen3-tts-flash.
  • 10 credits: qwen3.7-max, qwen3.7-max-preview, qwen3.6-max-preview, qwen3-max, qwen3-max-preview, qwen-max, qwen3.6-plus, qwen-plus, qwen-plus-latest, qwen3.5-omni-plus, qvq-max.
  • 5 credits: qwen3.5-plus, qwen3.6-flash, qwen-flash, qwq-plus, qwen3.5-omni-flash, qwen3-vl-plus, qwen-vl-max, text-embedding-v3, text-embedding-v4.
  • 1 credit: qwen3.7-plus, qwen3.5-flash, qwen-turbo, qwen-long-latest, qwen-long-2025-01-25, qwen3-vl-flash, qwen-vl-plus, qwen3-rerank.
Models not listed above are not available through AI credits. Add your own API keys to use them.

​
Check Your AI Credits

Check your remaining AI credits below your workspace name. Click the balance to open Integrations > Model Provider for usage details. The number of included AI credits depends on your plan:
After upgrading to Professional or Team, your new plan’s AI credits are used before any remaining Sandbox credits. The two balances cannot be combined for one model run.

​
Set Usage Priority

For a provider that supports AI credits, the first API key you add is automatically used before AI credits. To change the order, go to Integrations > Model Provider, click Configure on the provider card, and set Usage Priority:
  • AI credits first: Requests use AI credits until they run out, then switch to the selected API key. If the remaining balance cannot cover the model’s full cost, change the priority to API Key.
  • API key first: Requests use the selected API key. If a request cannot be sent with that key, it uses available AI credits.
If a request has already been sent with your API key but fails, it is not retried with AI credits. Fix the API key or provider issue, or change the priority to AI credits before retrying.
Last modified on August 24, 2026