GPT-3.5 Turbo Instruct API pricing
GPT-3.5 Turbo Instruct is an earlier OpenAI model, still sold. It costs $1.50 per million tokens it reads and $2 per million it writes, $1.63 on a mix of 3 read for each 1 written. OpenAI's newest models are GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna. Below are every price OpenAI lists for it, and what 8 everyday jobs cost a month on it.
- Input
- $1.50 per million tokens read
- Output
- $2 per million tokens written
- On everyday work
- $1.63 per million, 3 read for each 1 written
- Context window
- 4K tokens at once
- Capability score
- No independent score yet
Every GPT-3.5 Turbo Instruct price, per million tokens
In US dollars, as OpenAI prints them on its price page.
| InputTokens the model reads | $1.50 |
|---|---|
| Cached inputRead again from the cache | Not listed |
| Cache writePutting text in the cache | Not listed |
| OutputTokens it writes, thinking included | $2 |
| Blend 3:13 tokens read for each 1 written | $1.63 |
| Batch inputAnswers collected later | Not listed |
| Batch output | Not listed |
| Context windowTokens it holds at once | 4K tokens |
| Longest answerTokens it writes in 1 reply | 4K tokens |
What 8 everyday jobs cost a month on GPT-3.5 Turbo Instruct
Each job at its usual number of runs a day, for 30 days, at the standard price. Open any job in the cost calculator to change its size, or to set it beside every other model.
| Job | A run | A month |
|---|---|---|
| A chat question300 in, 500 out, 100 a day | 0.14 cents | $4.35 |
| A hard question, with thinking300 in, 1,400 out, 100 a day | 0.33 cents | $9.75 |
| A long report, summarised60,000 in, 2,000 out, 20 a day | Too long for its window | |
| A 40-turn agent task2,800,000 in, 40,000 out, 10 a day | Too long for its window | |
| A support bot's reply4,000 in, 300 out, 2,000 a day | Too long for its window | |
| Sort a message into a category250 in, 5 out, 50,000 a day | 0.039 cents | $578 |
| Read an invoice into fields1,500 in, 250 out, 1,000 a day | 0.27 cents | $82.50 |
| Review a pull request25,000 in, 4,500 out, 50 a day | Too long for its window | |
Compare GPT-3.5 Turbo Instruct
Both price lists side by side, and what the same 8 jobs cost on each.
Put this price on your own site
Writing about GPT-3.5 Turbo Instruct? Copy the code into your post, docs or course page and show its price as a small card. The card reads the same prices as this page, so when GPT-3.5 Turbo Instruct's price changes here it changes on your page too, and you never publish an old figure. It is free and needs no account.
The short script at the end lets the frame take the card's height. Where a page cannot take scripts, choose a fixed height.
<iframe src="https://tecyfy.com/embed/model/gpt-3-5-turbo-instruct" title="GPT-3.5 Turbo Instruct API price" width="100%" height="290" style="border:0;display:block" loading="lazy" allow="clipboard-write"></iframe>
<p><a href="https://tecyfy.com/prices/gpt-3-5-turbo-instruct">GPT-3.5 Turbo Instruct API price</a> by Tecyfy</p>
<script>addEventListener("message",function(e){if(e.origin!=="https://tecyfy.com"||!e.data||e.data.type!=="tecyfy:height")return;var h=Number(e.data.height);if(!(h>0))return;document.querySelectorAll("iframe").forEach(function(f){if(f.contentWindow===e.source)f.style.height=Math.min(h,20000)+"px"})})</script>How to read a model's prices
Every maker prints 2 main prices: one for the tokens a model reads (input) and one for the tokens it writes (output), each per million tokens. Writing costs more on almost every model, so a job that reads a long document and writes a line pays mostly the input price, and a chat that writes paragraphs pays mostly the output price. The blend, 3 tokens read for each 1 written, is one figure for ranking models on everyday work.
The other prices lower or raise the bill. Text read again from a cache costs a fraction of the input price, so a long system prompt or document sent with every request gets cheaper after the first. Work that can wait a day costs about half on a batch where the maker offers one. Some models charge more for every token once a prompt passes a set length.
Thinking is billed as output, and the same text is not the same number of tokens on every model, so the price per token is only the start. Each job below is priced the way the cost calculator prices it, at its usual runs a day for 30 days; open one to change its size, or compare the model with another side by side.
Questions
- How much does GPT-3.5 Turbo Instruct cost?
- GPT-3.5 Turbo Instruct costs $1.50 per million input tokens, the text it reads, and $2 per million output tokens, the text it writes, thinking included. On a mix of 3 tokens read for each 1 written, that is $1.63 per million.
- Does GPT-3.5 Turbo Instruct have a batch price?
- Not on OpenAI's price page, which lists no batch price for GPT-3.5 Turbo Instruct. The cache still lowers the price of text it reads again.
- What is the context window of GPT-3.5 Turbo Instruct?
- GPT-3.5 Turbo Instruct holds 4K tokens at once, about 6.1 pages of English text, and writes up to 4K tokens in one answer. The window holds your instructions, the files you attach, the conversation so far and the answer.
- How good is GPT-3.5 Turbo Instruct?
- Epoch AI has no independent score for GPT-3.5 Turbo Instruct yet. Run the same 10 tasks of your own on it and on a model you know, and compare the answers and the bills.
- Should I still use GPT-3.5 Turbo Instruct?
- It is still sold, so it keeps working, but OpenAI's newest models are GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna. Compare its price with theirs below, and run your own tasks on the newer model before you move.
Sources