AI Prompt Cost Calculator

Through the API, one typical prompt (500 tokens in, 500 tokens out — a question and a few paragraphs of answer) costs between $0.000300 on GPT-6 Luna and $0.0300 on Claude Fable 5.1, i.e. 0.03¢ to 3¢. A mid-priced model like Grok 4.7 costs $0.0040, so $1 buys about 250 prompts. Paste your own prompt above for an exact figure.

If you use the chat apps on a free or monthly plan, you don't pay per prompt — the subscription covers it. Developers calling the API pay per token: one typical prompt on GPT-6 Astra costs $0.0300, Claude Fable 5.1 costs $0.0300, Gemini 3.1 Pro costs $0.0070. Smaller models from the same providers cost a fraction of that.

About a quarter to a third of a watt-hour for a typical text prompt, according to the companies' own figures: Google reports 0.24 Wh for the median Gemini Apps text prompt; OpenAI (Sam Altman) reports 0.34 Wh for the average ChatGPT query; Microsoft reports 0.31 Wh for the typical query to frontier-scale LLMs (modeled). Google says its median Gemini prompt uses less energy than watching TV for nine seconds and about five drops of water. Long, reasoning-heavy or image prompts use more.

Through the API, one standard AI image costs between 1¢ and 13¢ depending on the model — usually more than a typical text prompt. See the full comparison on the AI image cost page.

Tokens are the units AI models read and write. A token can be a word, part of a word, or a single character; in English, 1 token is roughly 4 characters or 0.75 words. Every API call is billed by the tokens you send (input) and the tokens the model generates (output).

Input tokens × the model's input price, plus expected output tokens × its output price, gives the cost per request. Multiply by requests per day, then by 30, for the monthly estimate. Some models switch to higher long-context rates for very large prompts; the calculator applies those automatically. All prices are per 1 million tokens from each provider's official pricing page.

As of October 2026, the cheapest current models for a typical prompt are GPT-6 Luna, Mistral Small 4, gpt-oss-120B — all under $0.000375 per prompt. Cheap models handle classification, extraction and simple chat well; complex reasoning and code usually need a more capable model.

Among current models, Claude Fable 5.1 is the most expensive at $10.00 input / $50.00 output per million tokens — $0.0300 for a typical prompt, about 100x the cheapest option.

It depends mostly on the model. At 1,000 conversations a day (about 2,000 input and 500 output tokens each), GPT-6 Luna costs about $13.50/month, Grok 4.7 about $210.00/month, and Claude Fable 5.1 about $1,350.00/month. Model choice is the biggest cost lever.

Output tokens are generated one at a time, each needing a full pass through the model, while input tokens are processed in parallel. That's why output typically costs 2–8x more than input.

We count tokens in your browser with OpenAI's o200k_base tokenizer — the one used by GPT-4o and the GPT-5 family — so counts for OpenAI models are close to exact. Other providers use their own tokenizers and may count the same text somewhat differently (Anthropic, for example, says its newer Claude models produce more tokens per text than older ones), so treat non-OpenAI figures as close estimates. Your text never leaves your device.

Every price was checked against the provider's official pricing page in October 2026. Models that have been retired or have an announced shutdown date are labelled on their pages. Promotional prices and peak/off-peak rates are noted where they apply.

For simple tasks — classification, extraction, reformatting — start with a budget model such as GPT-6 Luna, Mistral Small 4. For everyday chat and coding, mid-priced models like Claude Haiku 4.5, Grok 4.7 are a good balance. For the hardest reasoning and agentic work, the frontier models (Claude Fable 5.1, GPT-6 Astra, Claude Opus 5.5) cost the most. Compare them at your real volume before committing.