> For the complete documentation index, see [llms.txt](https://docs.livecaller.io/livecaller/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.livecaller.io/livecaller/services/ai-assistant/ai-credits-payment-model.md).

# AI credits payment model

**How the system works:**\
**Credits as AI currency**\
1 credit = 1 USD.\
Every AI action consumes a certain number of credits.

**Factors and parameters affecting credits and cost:**

* **LLM Model:** Gemini 2.5 Flash, GPT 5.0, Claude Sonnet 4.5, etc. The more powerful the model, the more credits are consumed.
  * Example: Gemini 2.5 Flash may spend around $0.01 per session, while GPT 5.0 may spend $0.04 per session.
* **LLM Model Instructions:** These are pre-written instructions used by the model automatically in every session. The more detailed the instructions for the assistant, the more credits are consumed.

```
You are the AI shopping assistant for a tech e-commerce website. 
Your role is to help customers with questions about smartphones and prices. Do not answer off-topic questions, always try to talk about smartphones.

Rules:
- Always give clear, friendly, and professional answers. 
- Answer in the same language the customer uses (Georgian or English).
- If you don’t know something (like real-time stock or prices), politely say so and suggest 
the customer check the website or contact support. 
- Keep answers short and to the point (2–5 sentences). 
- Do not invent product details or prices that are not provided. 
- If a customer asks something not related to the shop (like history of iPhones), 
you may answer briefly (Really short answer), but always guide them back to shopping context.
- Avoid unnecessary long texts, use concise explanations.
```

**Input Tokens** – The length of the user’s message.

* Example: A longer message consumes more tokens, and therefore more credits.

**Output Tokens** – The length of the AI-generated response.

* Example: Longer responses consume more credits.

**Knowledge Base Size** –

* Example: This can be a short description of the company and frequently asked questions, or for e-commerce, a full product database with quantities, prices, etc.
* The larger the knowledge base, the more tokens are used, which affects credit consumption.

**Example of Usage – How the Price is Calculated**

**Scenario:**

* An organization with 1000 chat sessions per month
* Each chat averages 5–10 user questions
* Selected model: Gemini 2.5 Flash (fast, balanced in cost and quality)
* Integrated knowledge base: contains FAQs, team info, products, and company description

**Result:**

* Each chat session consumes approximately 0.10–0.15 credits (i.e., $0.10–$0.15)
* Monthly cost for 1000 chats: approximately $100–$150
* This price is optimized and efficient using modern AI

**Higher Intelligence Model: GPT-5**

* GPT-5 consumes more credits
* For the same volume of 1000 chats, the estimated cost with GPT-5 would be $300–$400 per month

**Transparency for the Organization:**

* The organization always sees:
  * How many credits remain
  * How many credits were used per session
  * Which model was selected and its cost
* Credits are consumed only if the digital assistant is activated

**Advantages:**

* Full control over expenses
* Flexible models – choose the AI level you need
* Transparent and easy-to-understand system