$0.09 for 100 English Edits: An Example Receipt

01 / A 100-edit receipt, with the assumptions visible

A writing assistant should make your English clearer without becoming another monthly bill. For someone drafting customer replies, job applications, or messages to colleagues, the useful question is simple: what would 100 edits cost?

Here is a worked example, not a CheapRelay price quote. Assume an input rate of $0.50 per million tokens and an output rate of $2.00 per million tokens. These are illustrative calculation rates, not advertised rates for a particular model. Check the current rates for your chosen model before spending.

• ■ Number of editing requests: 100.
• ■ Average input per request: 800 tokens, including instructions, the draft, and any context.
• ■ Average output per request: 250 tokens, including the rewrite and a short explanation.
• ■ Input receipt: 100 × 800 = 80,000 tokens; 80,000 ÷ 1,000,000 × $0.50 = $0.04.
• ■ Output receipt: 100 × 250 = 25,000 tokens; 25,000 ÷ 1,000,000 × $2.00 = $0.05.
• ■ Example total: $0.09 for 100 edits, or $0.0009 per edit.

That total assumes no separately billed cache activity or other charges. It is arithmetic, not a measured usage result. Token counts vary with text length, language, instructions, and the amount of explanation requested.

02 / Spend the tokens on meaning, not fancy English

For a non-native English speaker, a polished sentence is not necessarily a better sentence. An assistant can accidentally strengthen a promise, soften a complaint, or replace familiar vocabulary with language you would never use.

Give it a narrow job: correct grammar, improve clarity, and preserve meaning. Specify the audience and tone. A message to a landlord needs different wording from an academic paragraph, but neither automatically needs elaborate vocabulary.

Try this instruction: “Edit my English for a workplace email. Keep my meaning, names, numbers, and level of certainty unchanged. Use plain English. Return one revised version and explain at most two important changes. If my meaning is ambiguous, ask instead of guessing.”

For example, “We can possibly deliver Friday” should not become “We will deliver Friday.” That is a business commitment, not a grammar correction. Review dates, amounts, promises, and negatives before sending.

03 / The second version has its own receipt line

At the same assumed token counts and rates, asking for a second complete edit on every request doubles the example bill from $0.09 to $0.18. Scaling the original workflow to 1,000 edits produces an estimated $0.90 bill.

Those figures only hold if average usage stays the same. A follow-up that includes the entire previous conversation can carry more input than the first request. A request for five alternative versions can also produce much more output.

• ■ Ask for one rewrite by default, not five.
• ■ Limit explanations to the changes you want to learn.
• ■ Send the relevant paragraph rather than an entire document.
• ■ Start a fresh request when earlier conversation is unnecessary.

04 / A small CheapRelay setup without a subscription

CheapRelay offers pay-as-you-go AI API access with no monthly fee or subscription. Its OpenAI-compatible base URL is https://api.cheapapirelay.com/v1. One key provides access to multiple model families, so an existing OpenAI SDK application can use the relay without a separate integration for every provider.

Sign up for CheapRelay, check the current input, output, and cache rates, and test ten representative drafts before estimating a larger workload. Billing follows real token usage, including cache reads and writes; do not assume repeated instructions are free.

05 / Keep a ten-draft ledger before scaling

Record input tokens, output tokens, billed cost, and whether each edit preserved your meaning. Calculate the average cost per acceptable edit, counting retries too. That small ledger gives you a more useful budget than any hypothetical receipt—and helps you choose an assistant that improves your writing without rewriting your intentions.