Understanding Refly Tokens
A guide to understanding what AI tokens are, how they are calculated, and how they are managed within Refly's different subscription plans.
Introduction
This guide will explain what AI tokens are, how they are calculated, and how they are managed within Refly's different subscription plans.
AI Tokens in Subscription Plans
Each subscription plan includes a certain amount of AI tokens per month. These tokens are used by the AI models to process and generate responses.
Available Plans
- Starter: 3 million monthly AI tokens included.
- Growth: 9 million monthly AI tokens included.
- Enterprise: 23 million monthly AI tokens included.
Additional Token Charges
If a user exceeds the included amount of tokens in the current billing period, extra tokens are charged based on the plan:
- Starter: Additional tokens are charged at €16 per million tokens.
- Growth: Additional tokens are charged at €14 per million tokens.
- Enterprise: Additional tokens are charged at €11 per million tokens.
What is a Token?
A token is a unit of measure used by AI models to process text. Different AI providers and models have their own definitions and calculations for tokens.
Token Calculation
- OpenAI (GPT models): A token is typically around 3-4 characters. This means that a word like "hello" would be approximately two tokens.
To understand more about how tokens are calculated, you can refer to the documentation from the AI provider: OpenAI Token Documentation
Token Usage by AI Model
The token usage varies depending on the AI model you use. Different models have different costs and efficiencies, represented by multipliers that affect how quickly your included tokens are consumed.
GPT-4o:
- Multiplier: 1
- Upside: Enhanced performance with better handling of complex and nuanced queries.
- Downside: Higher operational cost.
GPT-4o Mini:
- Multiplier: 0.1
- Upside: Cost-effective with improved performance over GPT-3.5.
- Downside: Still less capable than the full GPT-4o model for the most complex queries.
Understanding how tokens are consumed based on the model can help you manage your usage more effectively and optimize your costs.
Example: If you are using the Starter plan and have used up your 3 million included tokens, additional tokens will be charged at €16 per million tokens. Using GPT-4o, each token counts as one Refly token, whereas with GPT-4o Mini, each token counts as 0.1 Refly tokens. Therefore, with GPT-4o Mini, you effectively get 10 times more processing from the same number of included tokens compared to using GPT-4o.
By understanding and managing your AI tokens, you can ensure that you are utilizing your plan effectively and controlling your costs.