Billing

Know more about Dify's billing plans to support expanding your usage.

Workspace-based Billing

The Dify platform has "workspaces" and "apps". A workspace can contain multiple apps. Each app has capabilities like prompt orchestration, LLM invocation, knowledge RAG, logging & annotation, and standard API delivery. We recommend one team or organization use one workspace, because our system bills on a per-workspace basis (calculated from total resource consumption within a workspace). For example:

Workspace 1  
App 1(Prompt, RAG, LLM, Knowledge base, Logging & Annotation, API)
App 2(Prompt, RAG, LLM, Knowledge base, Logging & Annotation, API) 
App 3(Prompt, RAG, LLM, Knowledge base, Logging & Annotation, API)
...
Workspace 2

Plan Quotas and Features

We offer a free plan for all users to test your AI app ideas, including 200 OpenAI model message calls. After using up the free allowance, you need to obtain LLM API keys from different model providers, and add them under Settings --> Model Providers to enable normal model capabilities.Upgrading your workspace to a paid plan means unlocking paid resources for that workspace. For example: upgrading to Professional allows creating over 10 apps (up to 50) with up to 200MB total vector storage quota combined across projects in that workspace. Different version quotas and features are as follows:

Metric	Sandbox	Professional	Team
pricing	Free	$59/month	$159/month
Model Providers	OpenAI,Anthropic,Azure OpenAI,Llama2,Hugging Face,Replicate	OpenAI,Anthropic,Azure OpenAI, Llama2,Hugging Face,Replicate	OpenAI,Anthropic,Azure OpenAI, Llama2,Hugging Face,Replicate
Team Members	1	3	Unlimited
Apps	10	50	Unlimited
Vector Storage	5MB	200MB	1GB
Document Processing Priority	Standard	Priority	Priority
Logo Change	/	/	√
Message Requests	500 per day	Unlimited	Unlimited
RAG API Requests Quota Limits	/	√ Coming soon	√ Coming soon
Annotation Quota Limits	10	2000	5000
Agent Model	/	√ Coming soon	√ Coming soon
Logs History	30 days	Unlimited	Unlimited

Check out the pricing page to learn more.

**Vector storage:**Vector storage is used to store knowledge bases as vectors for LLMs to understand. Each 1MB can store about 1.2million characters of vectorized data(estimated using OpenAI Embeddings, varies across models). How much the data shrinks depends on complexity and repetition in the content.

**Annotation Quota Limits:**Manual editing and annotation of responses provides customizable high-quality question-answering abilities for apps.

**Message Requests:**The number of times the Dify API is called daily during application sessions (rather than LLM API resource usage). It includes all messages generated from your applications via API calls or during WebApp sessions. Note:Daily quotas are refreshed at midnight Pacific Standard Time.

**RAG API Requests:**Refers to the number of API calls invoking only the knowledge base processing capabilities of Dify.

Monitor Resource Usage

You can view capacity usage details on your workspace's Billing page.

monitor resource usage

FAQ

1. What happens if my resource usage exceeds the Free plan before I upgrade to a paid plan?

   During Dify's Beta stage, excess quotas were provided for free to seed users' workspaces. After Dify's billing system goes live, your existing data will not be lost, but your workspace can no longer process additional text vectorization storage. You will need to upgrade to a suitable plan to continue using Dify.

2. What if neither the Professional nor Team plans meet my usage needs?

   If you are a large enterprise requiring more advanced plans, please email us at [e-mail].

3. Under what circumstances do I need to pay when using the CE version?

   When using the CE version, please follow our open source license terms. If you need commercial use, such as removing Dify's logo or requiring multiple workspaces, using Dify in a SaaS model, you will need to contact us at [e-mail] for commercial licensing.