How do messages per month work?

GPT-3.5-Turbo, the default chatbot model, costs 1 message credit per response. The more advanced GPT-4 model costs 20 credits per response due to its higher computational cost. You can select which AI model to use in the chatbot settings. GPT-4 requires 15 times more prompt tokens and 30 times more completion tokens from OpenAI than GPT-3.5-Turbo, which is why each response with the more capable model costs significantly more credits.