Reasoning

Our Fair Usage Policy limits prompts per time frame to allow all users access to all models without a single user negatively impacting the service for others by monopolizing model capacities. It is designed to prevent abuse and ensure reliability and availability for the overall user base.

Implementation

If a user exceeds the allowed limits, they can switch to a different model. However, 99% of all users never hit this limit, and the limit is only temporary.

As LLMs have different prices and different demands, we structured them into different categories:

  • Category 1 includes smaller, faster models. They allow unlimited requests, although there is still spam protection against abuse.
  • Category 2 models can receive 200 messages in three hours.
  • Category 3 models allow 100 messages in three hours.

Overview

Category 1: Unlimited (with spam protection)

  • GPT-3.5
  • GPT-4o mini
  • Claude 3 Haiku
  • Gemini 1.5 Flash

Category 2: 200 messages / 3 hours

  • GPT-4o
  • Claude 3.5 Sonnet
  • All Mistral models
  • All Llama models

Category 3: 100 messages / 3 hours

  • GPT-4 Turbo
  • Claude 3 Opus
  • Gemini 1.5 Pro