What Are Tokens? LLM Pricing and Limits Explained
Every LLM price, limit and speed figure is quoted in tokens. Understanding them takes five minutes and makes every AI budget conversation clearer.
- Published
What a token is
Models do not read words or letters; they read tokens — chunks of text from a fixed vocabulary. Common English words are often one token; rare words split into several. As a rough rule for English, 100 tokens is about 75 words.
Why it matters
- Pricing: you pay per input token (what you send) and per output token (what the model writes), with output usually priced higher.
- Limits: the context window — how much the model can consider at once — is measured in tokens.
- Speed: output is generated token by token, so long answers take longer.
The Indian-language factor
Tokenisers are often trained mostly on English. Text in Hindi, Marathi or other Indian scripts can take noticeably more tokens for the same meaning, which raises cost and uses more context. Measure on your real content before budgeting, and compare models — tokenisers differ.
Controlling token use
- Trim instructions to what changes behaviour.
- Retrieve fewer, better passages.
- Summarise long histories.
- Ask for concise answers.
- Cache the stable prefix.
Frequently asked questions
How do I count tokens?
Providers offer tokenizer tools and return token counts with every API response. Log them.
Are images and audio tokens too?
Yes — multimodal inputs are converted to tokens and billed accordingly.
Why is output more expensive?
Generating each output token requires a full model step, while input is processed in parallel.
