Enter your usage — get instant monthly cost estimates for GPT-4o, Claude, Gemini, Mistral, and more.
| Model | Input $/1M | Output $/1M | Cost for your usage | Relative cost |
|---|
Every LLM API charges per token — a unit roughly equal to 4 characters or 0.75 words. You pay separately for input tokens (your prompt + context) and output tokens (the model's reply). Prices are quoted per 1 million tokens.
| Model | Input $/1M | Output $/1M |
|---|---|---|
| GPT-4o | $2.50 | $10.00 |
| GPT-4o-mini | $0.15 | $0.60 |
| Claude 3.5 Sonnet | $3.00 | $15.00 |
| Claude 3.5 Haiku | $0.80 | $4.00 |
| Gemini 1.5 Pro | $1.25 | $5.00 |
| Gemini 1.5 Flash | $0.075 | $0.30 |
| Mistral Large | $2.00 | $6.00 |
| Mistral Small | $0.10 | $0.30 |
Output tokens are always 2–5× more expensive than input tokens. For chatbots and agents, outputs dominate cost. For classification or extraction tasks where you pass long documents and get short answers, input tokens dominate. This calculator accounts for both separately.
The biggest levers: (1) Use a smaller model for simple tasks — GPT-4o-mini at $0.15/1M vs GPT-4o at $2.50/1M is a 16× cost difference with similar quality for most tasks. (2) Shorten your system prompts — every token in every request costs money. (3) Use caching — Claude and OpenAI both offer prompt caching that can reduce costs 50-90% for repeated context.
Cost = (input_tokens × input_price_per_1M / 1,000,000) + (output_tokens × output_price_per_1M / 1,000,000). This calculator multiplies by your daily request count and the selected period (day/month/year).
For high-volume applications: Gemini 1.5 Flash ($0.075 input / $0.30 output per 1M) and Mistral Small ($0.10/$0.30) are the cheapest capable models. For quality-critical use cases: GPT-4o-mini offers excellent quality at $0.15/$0.60 per 1M tokens.
A short user message: ~50-150 tokens. A detailed prompt with instructions: 300-800 tokens. A long document for analysis: 2,000-8,000 tokens. A typical response: 200-500 tokens. Use our token counter to measure your specific prompts.
Yes. Every token in the conversation history counts as input tokens on each request. A chatbot that carries 10 messages of context before each reply might send 2,000 tokens of history + 100 new tokens — the history dominates the cost. This is why context management is critical for production apps.
In 2026, many teams still underestimate prompt caching. If your app reuses long system prompts, tool schemas, or retrieval headers, providers may bill cached input tokens far cheaper than uncached ones—but only when the prefix matches exactly. A common mistake is injecting timestamps, request IDs, or reordered JSON near the top of the prompt, which breaks cache hits and silently doubles costs. In this calculator, model both cached and uncached input shares separately before estimating monthly spend.