Two bills, one scenario
Make's credits page explains the split. For third-party AI modules that need your own connection, you pay Make for credits based on operations and your AI provider for tokens [1]. For non-AI apps, one operation equals one credit [1]. Make's OpenAI app page says the connection uses your API key and organization ID, and that you must purchase credits on the OpenAI billing page [2].
There is a second route. Make's OpenAI app includes a Simple text prompt module that works without an OpenAI account [2], and Make's credits page says Make's AI Provider uses credits based on tokens and operations, with a varying number of credits per run [1]. Make's credits page does publish token rates for its AI Provider, for example 3,616 input tokens or 452 output tokens per credit on its GPT-5 mini tier [1], but it describes credits as varying with the tokens consumed and the model selected. This guide prices the bring-your-own-key route; if you use Make's provider, measure credits per run in the run history instead.
The example scenario and its assumptions
A new row in a Google Sheet holds a block of text (meeting notes or a customer message). Make sends it to an OpenAI module with a summarizing prompt, then emails the summary. Every number below is an assumption chosen to be easy to replace.
| Assumption | Value | Why |
|---|---|---|
| Runs per month | 600 (about 20 a day) | Illustrative small-team volume |
| Input tokens per run | 2,000 | Prompt plus a long note; measure yours |
| Visible output tokens | 300 | A short summary |
| Reasoning tokens | 700 | Placeholder for reasoning models; billed as output [3] |
| Make modules per item | 3: Watch New Rows, OpenAI, Gmail | Trigger + 2 actions |
| Trigger schedule | Once an hour, all month | Polling cost shown separately |
| Make credit price | $0.0009 | Make's own example: a $9 plan with 10,000 credits [4] |
OpenAI cost per run and per month
OpenAI's pricing page lists standard prices per 1 million tokens [5]. Cost per run = input tokens x input price / 1,000,000 + (visible output + reasoning tokens) x output price / 1,000,000. For gpt-4o-mini, which is not a reasoning model, we count only the 300 visible output tokens.
| Model | Input / output per 1M tokens [5] | Cost per run | 600 runs / month |
|---|---|---|---|
| gpt-5-nano | $0.05 / $0.40 | $0.0001 + $0.0004 = $0.0005 | $0.30 |
| gpt-4o-mini | $0.15 / $0.60 | $0.0003 + $0.00018 = $0.00048 | About $0.29 |
| gpt-5-mini | $0.25 / $2.00 | $0.0005 + $0.002 = $0.0025 | $1.50 |
| gpt-5.4-mini | $0.75 / $4.50 | $0.0015 + $0.0045 = $0.006 | $3.60 |
| gpt-5.5 (under 272K context) | $5.00 / $30.00 | $0.01 + $0.03 = $0.04 | $24.00 |
Make credits per run and per month
Make's operations page says a trigger runs once per check regardless of how many bundles it returns, and each later module runs once per bundle [6]. With hourly polling, the trigger alone is 24 x 30 = 720 credits a month. Each processed row then adds 2 credits (OpenAI and Gmail), so 600 rows add 1,200. The total is about 1,920 credits a month, which exceeds Make Free's 1,000 monthly credits [7].
Using Make's own example rate of $0.0009 per credit [4], 1,920 credits is about $1.73 of plan value. If you run out and buy extra credits, Make says those include a 25% additional cost [4]. Combined with gpt-5-mini, the example costs roughly $1.50 in OpenAI tokens plus about 1,920 Make credits a month.
| Change | Make credits / month | OpenAI (gpt-5-mini) / month |
|---|---|---|
| Baseline: hourly polling, 600 runs | 720 + 1,200 = 1,920 | $1.50 |
| Poll every 4 hours instead | 180 + 1,200 = 1,380 | $1.50 |
| Batch 20 rows into one daily summary with an aggregator | About 4 per day x 30 = about 120 | Fewer calls, but each has more input |
| Cut reasoning tokens to 200 with lower effort | 1,920 | $0.0015 x 600 = $0.90 |
Where estimates go wrong
- Reasoning tokens: OpenAI bills them as output and says max_output_tokens includes them; if the limit is reached the response can be incomplete before any visible text [3]. Set a limit, and check reasoning_tokens in the usage data.
- Effort settings: OpenAI says lower reasoning effort favors speed and lower token usage, with supported values varying by model [3]. Test whether a summary is still good at low effort.
- Long inputs: input cost scales with text length. Trim signatures, quoted replies and boilerplate before the OpenAI module.
- Retries: a failed run that is retried repeats both the Make operations and the OpenAI call.
- Price changes: OpenAI lists many models at different prices [5]; record the date you checked and rerun the calculation when you change models.
- Make's AI provider route: dynamic credits per run [1], so measure rather than reuse this table.
Recommendation
For summaries that a person will read before acting, start with a small model such as gpt-5-mini or gpt-5-nano, cap output tokens, and spend your effort on the Make side: slower polling or batching changes the bill more than the model choice in this example. Move to a larger model only if a sample of summaries shows real errors, and keep a person in the loop for anything sent to customers.
Questions before you decide
Does Make charge extra for OpenAI modules?
Make's credits page says that for AI apps you connect with your own account, you pay Make for credits based on operations and your AI provider for tokens. Some built-in AI options, such as Simple text prompt through Make's AI provider, use dynamic credits based on tokens and other factors instead.
Are reasoning tokens billed?
Yes. OpenAI's reasoning guide says reasoning tokens are not visible through the API but are billed as output tokens and occupy space in the context window. The usage data reports them as reasoning_tokens.
How do I get real token numbers instead of assumptions?
Run the scenario on five typical inputs and read the usage figures OpenAI returns in each response (input tokens, output tokens and reasoning tokens). Replace the assumptions in the calculator with the averages.
Sources & verification
Product details and prices can change. Check the linked provider before buying.
- Make Help Center: Credits Accessed 29 Oct 2026
- Make apps documentation: OpenAI (ChatGPT, Whisper) Accessed 29 Oct 2026
- OpenAI API guide: Reasoning models Accessed 29 Oct 2026
- Make Help Center: Extra credits Accessed 29 Oct 2026
- OpenAI API pricing Accessed 29 Oct 2026
- Make Help Center: Operations Accessed 29 Oct 2026
- Make pricing and billing FAQ Accessed 29 Oct 2026
Sources link directly to providers. Product buttons may use separately labeled affiliate links. Read our disclosure.