AI Token Cost Basics: Input, Output, and Workloads
Token pricing is multiplication with two rates: prompt (input) tokens and completion (output) tokens. This guide explains the workload math behind our token and LLM API calculators without claiming a live price feed.
What a token roughly is
Providers bill on tokens—subword pieces of text—not whole words. Exact tokenization differs by model. For planning, use measured token counts from logs or a tokenizer, then apply the rate card you copied from the provider.
Price one workload
Count input tokens
Prompt, system message, and retrieved context.
Estimate output tokens
Use average completion length from logs when possible.
Apply separate rates
Cost ≈ input_tokens × input_rate + output_tokens × output_rate.
Scale carefully
Multiply by requests only after one-shot cost looks right.
When to switch tools
Use the token calculator for a known single workload. Use the LLM API budget page for recurring monthly request volume plus fixed costs. Training and video tools price accelerators and minutes—not tokens.