AI API Cost Calculator

Almost every large language model API is billed by the token rather than by the request. A token is a chunk of text — roughly four characters or about three-quarters of an English word — and both the text you send (the input, or prompt) and the text the model generates (the output, or completion) are counted. Providers quote prices per million tokens, which makes headline numbers look small until you multiply them by the volume of a real application. This calculator turns those per-million rates into the numbers that actually matter: cost per request, and cost across thousands or millions of requests.

The single most important thing to understand is that input and output tokens are priced differently. Generating text is far more compute-intensive than reading it, so output almost always costs several times more than input — commonly three to five times. A model advertised at a cheap input rate can still be expensive if your workload produces long, verbose answers. When you compare models in the table above, watch the output column closely, because for chat and generation workloads it usually dominates the bill.

The cheapest model is rarely the right default, and the most expensive one is rarely necessary. A practical approach is to route by task difficulty: use a small, low-cost model (such as a mini or nano tier, Gemini Flash-Lite, or DeepSeek) for classification, extraction, and simple summarisation, and reserve a flagship model (GPT-5.5, Claude Opus, Gemini Pro) for complex reasoning, long-form generation, or high-stakes output. Because the price gap between tiers is often 10× to 50×, moving routine traffic to a cheaper model is usually the biggest single lever on an AI bill.

Estimate before you build, not after. Plug your realistic average token counts and monthly request volume into the calculator, compare the top few candidates in the table, and you will often find that two models with similar quality differ enormously in cost. For high-volume or non-urgent workloads, also check whether a provider offers a batch tier (frequently around a 50% discount) — those rates are not shown here but are listed on each provider's official pricing page linked above.

Disclaimer: This AI API Cost Calculator is provided for general estimation and educational purposes. Prices were compiled from official provider pages in July 2026and can change at any time; regional endpoints, data-residency options, batch and priority tiers, fine-tuning, and server-side tools (such as web search) may carry different or additional charges. INR figures are approximate conversions from USD, which is the billing currency for these APIs. Always confirm current pricing on the provider's official page before making purchasing or architectural decisions.