Tokens and Context Windows

Tokens

Models don't read characters or words; they read tokens, pieces of text from a fixed vocabulary. In English, one token is roughly 4 characters or three quarters of a word; code, other languages and unusual words use more tokens.

• Input tokens: everything you send (system prompt, conversation, documents, tool definitions).
• Output tokens: what the model generates.
• Pricing is per million input and output tokens; output tokens usually cost more.

The context window

The context window is the maximum number of tokens (input plus output) the model can handle in one request. Everything the model "knows" about your request must fit in it; the model has no memory between requests unless you send the history again.

def estimate_tokens(text):
    return max(1, len(text) // 4)          # rough rule for English text

system = "You are a helpful support assistant for an online bookshop."
history = ["Where is my order 10042?", "It shipped yesterday and arrives Friday."] * 20
question = "Can I change the delivery address?"
total = estimate_tokens(system) + sum(estimate_tokens(m) for m in history) + estimate_tokens(question)
print("estimated input tokens:", total)
price_in, price_out = 3.00, 15.00          # illustrative $ per million tokens
print(f"cost of one call with 300 output tokens: ${(total * price_in + 300 * price_out) / 1e6:.5f}")

estimated input tokens: 342
cost of one call with 300 output tokens: $0.00553

For exact counts, model APIs provide token-counting endpoints; the 4-characters rule is only an estimate. See Context windows and Token counting.