Back to glossary

Term

Output Token

Tokens an AI model produces as its response. Billed separately and usually three to five times more expensive than input tokens because the model has to actively generate them.

Output token — in more detail

Output tokens are everything the model produces as a response — generated token by token via billions of probability calculations. That active generation step is what makes output more expensive than input: for every single output token the model has to push the entire context through its layers again. Typical price ranges are $0.60 to $15 per million output tokens — several times the input price.

Example / practical context

With Claude Opus at $75/MTok output, a 2,000-token answer costs $0.15 — fine for a short reply, but it adds up quickly for longer ones. Reasoning models additionally produce “invisible” output tokens (thinking tokens) that are not displayed but still fully billed — a single reasoning answer can easily consume 5,000–20,000 output tokens. To keep costs in check, set max_tokens as a hard cap per request.

Input tokens are what gets billed before generation — cheaper because they are only read, not produced. Streaming does not change the price; it just delivers the output tokens piece by piece instead of all at once. Cached output tokens are not offered by most providers — caching only works on the input side.

Entdecke mehr