Token
The atomic unit a language model reads and writes - usually a sub-word piece used for pricing and limits.
- Common English word ≈ 1 token; rare words and code use more.
- Rule of thumb: 1000 tokens ≈ 750 English words.
- One token per generation step.
When to use: When estimating cost, latency, or context budget for any LLM workload.
Example: GPT-5: ~$1.25 per million input tokens, $10 per million output tokens (Aug 2025).