Cost per Token
Biaya per Token
What you pay for each unit of text the model reads or writes.
Providers bill input tokens and output tokens separately, and output is usually several times more expensive per token because generating a token costs more than reading one. Prices are listed per million tokens, which makes small comparisons misleading.
The bill follows a simple formula: requests times tokens per request times price per token, plus whatever the tools and extra calls add. A system that works perfectly on ten requests can become expensive at ten thousand, because the multiplier is volume.
This is why the cost of an AI feature is often dominated by something nobody planned: the retry, the verbose system prompt, the tool definition loaded on every call, the agent loop that runs fifteen steps. Cutting those is usually a bigger win than switching to a cheaper model.
Related terms
Token
The unit a model reads and writes. Roughly word-sized, but not exactly.
Model Provider
The company that serves a model to you. Usually someone else's server.
Caching
Storing results so the same work is not repeated, and the cost drops with it.
Latency
How long you wait. The time to the first token matters more than the total.
Quantization
Storing weights in fewer bits so a model runs with less memory and cost.
Want this working inside your team?
It starts with one conversation about your workflow.