Understanding token usage

Last updated: 2026-06-10

Your lm9 LLM plan includes a monthly token allowance. This page explains what tokens are, what counts toward your limit, and how to get the most from your plan. Plan prices and allowances are on our pricing page.

What is a token?

Tokens are small pieces of text the AI model reads and writes — roughly a word or part of a word. We measure usage in tokens because that is how the underlying models bill compute. Your monthly allowance is a cap on how many tokens you can use across chat and API access in a billing period.

Rough guide for English: about 4 characters ≈ 1 token. A short question might use a few hundred tokens total; a long conversation or large pasted document can use thousands or more in a single request.

How we count your usage

Each AI request uses two kinds of tokens:

• Input tokens — everything sent to the model: your messages, system instructions, conversation history, and any document or search context included in the prompt.
• Output tokens — the model’s reply.

Your usage = input tokens + output tokens, added together across all qualifying requests in the month.

Usage is counted when you:

• Chat at llm.lm9.org/chat
• Use the API token (Free, Pro, and Power) — same monthly pool as web chat
• Build or query a knowledge base (RAG), including embedding documents for search

Counts come from the model runtime after each successful response. Your balance resets on the first day of each calendar month at 00:00 UTC (not necessarily your local timezone or subscription anniversary date).

Monthly allowances by plan

Token limits were increased by 50% in June 2026. Current monthly allowances:

• Trial (15 days): 5,000,000 tokens
• Starter: 50,000,000 tokens / month
• Pro: 300,000,000 tokens / month
• Power: 1,000,000,000 tokens / month

When you reach your monthly limit, new chat and API requests pause until the next UTC month or until you upgrade your plan. If that happens, you will see a message showing how many tokens you have used.

What uses more tokens?

• Longer chats — each new message resends earlier messages in that thread, so cost grows as the conversation continues.
• Large pasted text or uploads — big inputs consume many tokens at once.
• Knowledge bases (RAG) — retrieved document snippets add to the prompt; indexing documents also uses tokens.
• Longer AI answers — detailed explanations, code, or step-by-step reasoning increase output tokens.
• Regenerating a response — each attempt is a new request (full input plus new output).
• Multiple chats or API sessions — all activity in the month adds to the same allowance.

Model size does not multiply tokens. A smaller model and a larger one use roughly the same token count for the same text. Plans differ mainly in which models you can access, speed, and priority — not in a separate per-model token multiplier.

What does not use LLM tokens?

• Signing in, managing your account, or browsing available models
• Image generation (where available on Pro and Power) — separate from your chat token allowance

Tips to use fewer tokens

• Start a new chat when the topic changes instead of keeping one very long thread.
• Keep system prompts focused — shorter instructions leave more room for answers.
• Avoid pasting huge documents when a summary or excerpt will do.
• Ask for concise replies when you do not need long-form output.

Questions?

See our Terms of Service for billing and trial details, or contact us if something looks wrong with your usage.