AI Usage Model

PumpTON leverages OpenAI's GPT models through API calls to build natural language-based conversational AI agents. In this context, a "token" represents the smallest unit of text that the AI understands. For example, a simple sentence like “Hello, how are you?” consists of about 6 tokens.

OpenAI’s pricing model charges based on the sum of input and output tokens per interaction. This means that both the prompt sent by the user and the response generated by the AI are counted toward billing.

PumpTON primarily supports GPT-3.5-turbo and GPT-4 for its agents. Additional open-source models such as Hugging Face and Mistral will be integrated in future updates.

• GPT-3.5: Approximately $0.0015 per 1K tokens (input + output combined)

• GPT-4: Approximately $0.06 per 1K tokens (input + output combined)

Based on these rates, PumpTON has implemented a pricing strategy that ensures a 30–40% profit margin per call. For example, a GPT-3.5 call costs roughly $0.00075 in average use cases. By setting the usage fee at 0.15 TON (about $0.18), PumpTON ensures profitability while covering operational expenses. For GPT-4, where the cost per call is around $0.03, the usage fee is set at 0.45 TON (about $0.54), maintaining a similar margin structure.