Have AI Draft a 1,500-Word Client Article: From Brief to File, and What It Costs by the Token

A client wants a 1,500-word article by Friday. You want a first draft in the client's voice that you can edit in an hour instead of writing in four. This page is that job, done by API or in a browser, from the brief you write to the file you send, and then what the draft cost in tokens. It reads the providers' own guides for the steps and their own pricing pages for the money; nothing here is a benchmark or a verdict on which model writes better, because none of them publishes one.

What you hand over

A 1,500-word draft in the client's register, with the structure they asked for, without a chatty opening line, and with every fact, name and figure checked by you. The model produces the first three; the fourth is your edit pass and it is what the client is paying for.

Step one: the brief is the system instruction

Every provider's guide says the same thing about where the brief goes: not in the message, in the instruction layer above it. OpenAI calls it the instructions parameter and says it gives the model high-level instructions on tone, goals and examples of correct responses, taking priority over the prompt in input. Anthropic calls it the system prompt and its best-practices page spells out what to put there: be clear and direct, give the model a role, wrap 3 to 5 examples in <example> tags so it can tell examples from instructions, and put long material, such as the client's previous articles, near the top of the prompt above the instructions. It adds one trick for source-heavy pieces: ask the model to quote the relevant parts of the documents first, before it writes, so it stays on the source.

So the brief for our article is one block of text: who the reader is, the length in words, the section list if there is one, the tone in a sentence, three things never to do, and three of the client's own paragraphs as examples. Write it once. Everything below reuses it.

Path A: in a browser, no code

Gemini's Google AI Studio opens the Playground with a new chat prompt. Click Run settings in the top-right corner to open the panel, paste the brief into the System Instructions field, type the topic in the message box and click Run. The quickstart's own example changes the system instruction mid-conversation and reruns, which is exactly the loop you want: tighten the brief, run again, until the register is right. When it is, Get code turns the prompt into API code in your language, and the prompt can be saved and shared.

Mistral's Studio starts in Free mode with no credit card. Under Build, Prompts, New prompt saves the brief as a template with a title and description; saving a change later creates a new version while the old one stays available for rollback, and Share keeps it private or opens it to a workspace. The Playground is where you test it. For a client you write for every month, that versioned prompt is the asset.

Remember the price of the free door on the first of these. Gemini's pricing page prints, under every model, that Free tier content is used to improve Google's products and Paid tier content is not. Fine for your own topic; not for the client's draft strategy pasted in as an example.

Path B: one request

OpenAI's text guide shows the whole call with its command-line client:

openai responses create --model "gpt-6-astra" --input "Write a one-sentence bedtime story about a unicorn." --raw-output --transform 'output.#(type=="message").content.0.text'

Swap the model for the one you priced below, add --instructions with the brief, and put the topic in --input. The guide also says to pin a dated model snapshot in production so behaviour does not shift under you between pieces; for a client's monthly column that is the difference between one voice and twelve.

Anthropic's first call is a POST to /v1/messages with model, max_tokens and a messages array; the system prompt is where the brief goes. One number in the getting-started example matters here: it sets max_tokens to 1000. A 1,500-word draft is about 2,000 tokens, so leave that at 1000 and the article stops mid-sentence. Set it above the draft. The response reports input_tokens and output_tokens, so every call tells you what it cost.

Mistral's first request goes to /v1/chat/completions with mistral-large-latest. The quickstart's error table is worth reading before you start: a 402 Payment Required means there is no payment method on the account, added under Admin, Subscriptions, Billing, and a 429 means you hit the Free mode rate limit and should wait.

The size of the job, in tokens

OpenAI's help page gives the conversion: 1 token is approximately 4 characters or three-quarters of a word, and 100 tokens are approximately 75 words, for English text. So 1,500 words out is about 2,000 tokens, and a 300-word brief in is about 400 tokens. Those two numbers are the whole bill.

What the draft costs

Every provider publishes two prices per million tokens, input and output, and for drafting the output price is the one that matters, because the draft is five times the brief. The chart above is the output column as published. Per million tokens out: Together AI lists DeepSeek V4 Flash at $0.28; OpenAI lists gpt-4o-mini at $0.60, gpt-5.6-luna at $1.20 and gpt-5.6-sol at $20.00 on promotional pricing through at least November 21, 2026; Mistral Large is $1.5; Gemini 3.8 Flash is $3.75 on the Paid tier until December 31, 2026 and $7.50 from January 1, 2027; Claude Haiku 4.5 is $5, Sonnet 5 $10, Opus 5 $25 and Fable 5.1 $50.

Multiply by 2,000 tokens. The draft costs a tenth of a cent on gpt-4o-mini and ten cents on Fable 5.1; the 400 input tokens add a fraction of that. The spread between the cheapest and dearest model on the page is under a dime per article. The AI job cost calculator does this for any length; the Claude and OpenAI compare reads the two biggest price lists side by side.

Getting through the door is also priced differently. OpenAI bills per token from the first call with no plan fee. Anthropic sells prepaid credits, pay-as-you-go, with no published minimum. Gemini's Free tier is free and its Paid tier is pay-as-you-go. Mistral Studio opens in Free mode.

The edit pass

The model's draft arrives with two habits the guides tell you how to remove. The first is the preamble. Anthropic's page gives the instruction verbatim: Respond directly without preamble, do not start with phrases like Here is or Based on. The second is length. The same page says Claude Opus 5 defaults to longer responses than earlier models and that changing the effort setting does not reliably shorten them, so ask for conciseness in the brief. On the OpenAI side, the snapshot pin is the fix for drift.

Then your pass, which no provider does for you: every number, name and claim against the client's sources, the section order against the brief, the register against the three example paragraphs. Save the result as the file the client opens and send it. That file is the only lasting copy, for a reason under the terms below.

Thirty drafts instead of one

If the client wants a month of posts, do not send thirty requests at noon. Anthropic's Message Batches API charges 50% of the standard API price; most batches finish in less than 1 hour, anything unfinished expires at 24 hours and is not billed, a batch holds up to 100,000 requests or 256 MB, and results stay downloadable for 29 days. Mistral's pricing page says batch processing halves its price as well. The same thirty drafts, half the bill, ready in the morning.

Who owns the draft

OpenAI's services agreement assigns to the customer all of its right, title and interest in the Output, and the customer keeps ownership of its Input. Anthropic's commercial terms say the customer owns the Outputs and keeps all rights to Inputs, that Anthropic assigns its interest in Outputs, and that it may not train models on Customer Content. Mistral's open-weight models are Apache 2.0 for research and individual use, with commercial deployments needing a Mistral licence. Gemini's line is the tier row above: Paid does not train, Free does.

What is deleted when you stop

Anthropic auto-deletes API inputs and outputs within 30 days, so the only lasting copy is the one you store. OpenAI deletes all Customer Content within thirty days of the agreement terminating and hands nothing back. OpenAI may limit or suspend access for a breach of the agreement or its policies, without prior notice where necessary; Anthropic may suspend all access over a suspected Usage Policy breach and disclaims liability for the loss that causes. None of this touches a draft you already sent to the client. All of it touches the one you left in the console.

What to do today

Take one piece you have already written for a client. Turn its brief into a system instruction with three of the client's paragraphs as examples. Run it once in AI Studio or Mistral Studio, once by API with max_tokens above 2,000, and time your edit pass on each. The token bill will be a few cents; the edit time is your real cost, and it is the number to quote from.