Skip to main content

AI product budgeting

How Much Does an AI Chatbot Cost to Run?

An AI chatbot budget combines model inference with retrieval, storage, observability, and operational overhead. Start with cost per conversation turn, then scale by active users and turns per user.

Formula-driven examplesSource-linked pricing snapshots

Model the conversation turn

Include the system prompt, conversation history, retrieved context, user message, and generated response. History growth can make later turns more expensive than the first.

Use requests per active user per day rather than total signups. An activation assumption that is too optimistic can distort both cost and revenue planning.

monthly model cost = cost per turn × turns per active user per day × active users per day × 30

Add non-model costs

Production chatbots may also pay for embeddings, vector search, reranking, moderation, transcription, image generation, logging, support, and retries. Keep these line items visible instead of hiding them in a single model-rate assumption.

Worked example

GPT-4.1 mini: 1M input tokens + 1M output tokens

Using the versioned rates below, this example workload is estimated at $2.00. This isolates provider usage only and does not include taxes, regional premiums, retries, storage, network traffic, or unrelated infrastructure.

Current pricing references

These versioned records support the examples above. Check the date and provider source before using them in a production forecast.

Provider / modelInput or unitOutputStatusSource

OpenAI

GPT-4.1 mini

$0.40 per 1M tokens$1.60 / 1MVerifiedOpenAI model pricing

Checked Jun 21, 2026

Google Gemini

Gemini 2.5 Flash

$0.30 per 1M tokens$2.50 / 1MVerifiedGoogle Gemini API pricing

Checked Jun 21, 2026

Anthropic

Claude Haiku 4.5

$1.00 per 1M tokens$5.00 / 1MVerifiedAnthropic Claude API pricing

Checked Jun 21, 2026

Frequently asked questions

What is the biggest chatbot cost driver?

At scale, generated output, repeated context, and request volume are common drivers. The dominant factor depends on the model and conversation design.

How much safety buffer should I add?

Use observed variance when available. Early planning often benefits from a modest buffer for longer responses, prompt wrappers, and traffic uncertainty.

Related calculators and guides