GPT-3.5 Turbo Instruct vs GPT-6 Luna
GPT-6 Luna is one of OpenAI's newest models and GPT-3.5 Turbo Instruct an earlier one, so this is also what moving up costs or saves. GPT-6 Luna is the cheaper of the 2: for the same tokens, 3 read for each 1 written, GPT-3.5 Turbo Instruct costs 8.1 times as much. Below are both price lists, what 8 everyday jobs cost a month on each, and how many tokens the same text takes on each, which can change the answer.
- For the same tokens
- GPT-6 Luna is cheaper; GPT-3.5 Turbo Instruct costs 8.1 times as much
- Cheaper to read
- GPT-6 Luna, $0.10 against $1.50
- Cheaper to write
- GPT-6 Luna, $0.50 against $2
- Scores higher
- Not known, GPT-3.5 Turbo Instruct has no independent score yet
- Larger window
- GPT-6 Luna, 1.05M against 4K tokens
Price per million tokens, side by side
In US dollars, as each maker prints them. The better figure in each row is green.
| GPT-3.5 Turbo InstructOpenAI · earlier model | GPT-6 LunaOpenAI | |
|---|---|---|
| InputTokens the model reads | $1.50 | $0.10 |
| Cached inputRead again from the cache | Not listed | $0.01 |
| Cache writePutting text in the cache | Not listed | $0.125 |
| OutputTokens it writes, thinking included | $2 | $0.50 |
| Blend 3:13 tokens read for each 1 written | $1.625 | $0.20 |
| Batch inputAnswers collected later | Not listed | $0.05 |
| Batch output | Not listed | $0.25 |
| Long prompt inputAbove 272K on GPT-6 Luna | $1.50 | $0.20 |
| Long prompt outputThe standard price where there is no long-prompt price | $2 | $0.75 |
| Context windowTokens it holds at once | 4K | 1.05M |
| Capability scoreEpoch Capabilities Index; GPT-5 is 150 | Not listed | 156 |
| Longest answerTokens it writes in 1 reply | 4K | 128K |
What 8 everyday jobs cost a month on each
Each job at its usual number of runs a day, for 30 days, at the standard price. The cheaper of the 2 is green; open any job in the calculator to change it.
| Job | GPT-3.5 Turbo Instruct | GPT-6 Luna | Difference |
|---|---|---|---|
| A chat question300 in, 500 out, 100 a day | $4.350.14 cents a run | 84 cents0.028 cents a run | 5.2 times |
| A hard question, with thinking300 in, 1,400 out, 100 a day | $9.750.33 cents a run | $2.190.073 cents a run | 4.5 times |
| A long report, summarised60,000 in, 2,000 out, 20 a day | Too long for its window9.4 cents a run | $4.200.7 cents a run | |
| A 40-turn agent task2,800,000 in, 40,000 out, 10 a day | Too long for its window$4.28 a run | $21.967.3 cents a run | |
| A support bot's reply4,000 in, 300 out, 2,000 a day | Too long for its window0.66 cents a run | $22.200.037 cents a run | |
| Sort a message into a category250 in, 5 out, 50,000 a day | $5780.039 cents a run | $41.250.0028 cents a run | 14 times |
| Read an invoice into fields1,500 in, 250 out, 1,000 a day | $82.500.27 cents a run | $8.250.027 cents a run | 10 times |
| Review a pull request25,000 in, 4,500 out, 50 a day | Too long for its window4.7 cents a run | $7.130.48 cents a run |
Your own job on both
Start from a job, then set its size, how often it runs and whether the answers can wait. Both bills follow every change.
A short question and a few paragraphs back.
about 230 words
Instructions or documents the model has read before cost far less the second time.
about 380 words
A month of this job
- GPT-3.5 Turbo Instruct$4.35
0.14 cents a run
- GPT-6 Luna84 cents
0.028 cents a run
GPT-6 Luna saves $3.51 a month; GPT-3.5 Turbo Instruct costs 5.2 times as much.
This job on every modelWhy the price per token is only the start
Every maker prints 2 main prices: one for the tokens a model reads (input) and one for the tokens it writes (output). On most current models writing costs 3 to 5 times as much as reading, so which model is cheaper can depend on the job. A job that reads a long document and writes a line pays mostly the input price; a chat that writes paragraphs pays mostly the output price. That is why each comparison prices 8 jobs and not one figure.
The same text is not the same number of tokens on every model. Each maker cuts text its own way, and Anthropic says its models from Claude 4.7 on produce about 30 percent more tokens for the same text than earlier ones. A model with the lower price per token can still send the larger bill, so each comparison counts one of our lessons the way both makers count it.
Discounts change the order too. Work that can wait costs half on a batch at the makers that offer one, and text read again from a cache costs a quarter of the input price or less on every current model here. The price table on each comparison shows who offers what.
Thinking is billed as output, and models differ in how much they think and how long they answer. A price list cannot show that. Before you switch, run the same 10 tasks on both models and compare the bills.
Questions
- Which is cheaper, GPT-3.5 Turbo Instruct or GPT-6 Luna?
- GPT-6 Luna is cheaper: $0.10 per million tokens read and $0.50 per million written, against $1.50 and $2 on GPT-3.5 Turbo Instruct. On a mix of 3 tokens read for each 1 written, GPT-3.5 Turbo Instruct costs 8.1 times as much.
- Which is better, GPT-3.5 Turbo Instruct or GPT-6 Luna?
- GPT-6 Luna scores 156 on the Epoch Capabilities Index, an independent score from Epoch AI that combines over 50 benchmarks (GPT-5 is 150). GPT-3.5 Turbo Instruct has no independent score yet, so the only fair test is to run the same 10 tasks on both.
- Which has the larger context window, GPT-3.5 Turbo Instruct or GPT-6 Luna?
- GPT-6 Luna, with 1.05M tokens against 4K on GPT-3.5 Turbo Instruct. The window holds your instructions, the files you attach, the conversation so far and the answer, all at once.
- Does the same text take the same number of tokens on GPT-3.5 Turbo Instruct and GPT-6 Luna?
- Yes, both are counted the same way: our lesson "How AI writes", 1,700 words long, is about 2,100 tokens on each.
- Is it cheaper to move from GPT-3.5 Turbo Instruct to GPT-6 Luna?
- Yes, per token: GPT-6 Luna costs $0.10 per million tokens read and $0.50 written, against $1.50 and $2 on GPT-3.5 Turbo Instruct, which costs 8.1 times as much for the same tokens. Run your own tasks on it before you switch, since answers, length and thinking can differ.
- Can I pay less on GPT-3.5 Turbo Instruct and GPT-6 Luna for work that can wait?
- On GPT-6 Luna only: with a batch, where you collect the answers later, it costs $0.05 per million tokens read and $0.25 written. OpenAI's price page lists no batch price for GPT-3.5 Turbo Instruct.
Sources