AI API Pricing
Calculator

Tokens are the fundamental units that AI language models use to process text. Roughly, 1 token equals about 4 characters or 0.75 words in English. AI providers charge per token, with separate rates for input tokens (your prompt) and output tokens (the model's response). Prices are quoted per 1 million tokens. For example, GPT-5.4 charges $2.50 per 1M input tokens and $15.00 per 1M output tokens.

As of March 2026, the cheapest AI APIs are Google's Gemini 3.1 Flash Lite at $0.25/$1.50 per 1M tokens (input/output), GPT-4.1 Nano at $0.10/$0.40, and Gemini 2.5 Flash at $0.30/$2.50. DeepSeek V3 is also very competitive at $0.27/$1.10. For the best value, use prompt caching (up to 90% savings) and batch processing (50% off).

Prompt caching stores frequently repeated parts of your prompts (like system instructions) so the API doesn't reprocess them each time. Cached input tokens cost significantly less: Anthropic offers 90% off, OpenAI's GPT-5.4 family offers 50% off (GPT-4.1 family 75% off), and Google offers 75% off. If your application sends similar prompts repeatedly, caching can dramatically reduce your API bill.

Batch API lets you submit large numbers of requests to be processed asynchronously (typically within 24 hours). In exchange for the longer wait time, providers offer 50% off both input and output token costs. Batch pricing is ideal for non-time-sensitive tasks like data processing, content generation, and bulk analysis.

To estimate monthly costs, multiply your average input tokens per request by the input price, add average output tokens times the output price, then multiply by daily request volume and 30 days. Our calculator does this automatically — just enter your tokens per request and daily volume to see costs across all models. Don't forget to factor in prompt caching and batch pricing if applicable.

GPT-5.4 ($2.50/$15.00 per 1M tokens) and Claude Sonnet 4.6 ($3.00/$15.00) are both flagship-to-mid tier models. GPT-5.4 has a larger 1.05M token context window vs. Sonnet's 200K, and is OpenAI's latest frontier model. Sonnet 4.6 is known for strong coding and reasoning. For cost-sensitive applications, GPT-4.1 ($2.00/$8.00) offers better value; for tasks requiring nuanced reasoning, Sonnet 4.6 may justify the premium.

OpenAI's o-series models (o3, o4-mini) use internal 'reasoning tokens' — the model thinks step-by-step before generating its final response. These reasoning tokens are billed as output tokens but aren't visible in the API response. This means the actual cost of o-series requests can be higher than expected, especially for complex reasoning tasks. Our calculator uses the standard output rates; your actual costs may be higher due to hidden reasoning tokens.

Our token counter uses an approximation of ~1 token per 4 characters, which is accurate within 10-15% for English text. Different models use different tokenizers (BPE, SentencePiece, etc.), so exact counts vary slightly. For precise token counts, use the official tokenizer tools from each provider. Our estimates are reliable enough for cost planning and budgeting purposes.