What are credits?
A plain-language introduction to ChatGPT's points-based usage pricing
Key takeaway
Credits are usage points used in ChatGPT's business plans. When you send a message, a quantity of credits is deducted according to the feature selected. Credits are another pricing unit, distinct from the token billing discussed earlier. This article describes the July 2026 rate-table snapshot cited below.
Credits are a little like energy points in a game
Think of a game showing “This action used 10 energy points.” In the business-plan pricing described here, requesting an AI task deducts credits according to the feature. Simple requests consume fewer credits; complex work consumes more.
Energy points in a game
Each action consumes an amount that depends on how demanding it is.
A staff cafeteria payment card
Rice, a set meal, and a special lunch box cost different amounts. The total depends on what you order.
A prepaid card
An organization receives a credit allowance, and the balance decreases as it is used.
Imagine a self-checkout. Each scan adds an amount to the bill. A cheap item (a light message) and an expensive item (a demanding feature) are both one scan, but their prices differ greatly. That is the idea behind credits.
Different models have different weights
Even when the action is simply asking ChatGPT a question, credit consumption differs greatly by model. Higher-capability models consume more credits per message in the rate table described here.
GPT-5.5 Instant (light and fast)
For everyday questions, email drafts, translations, and summaries. Designed for common tasks. Listed as unlimited in this snapshot.
GPT-5.5 Thinking (careful reasoning)
For complex documents, logical analysis, and comparisons from several perspectives. 10 credits per request.
GPT-5.5 Pro (most demanding, highest quality)
For tasks requiring advanced expertise, precise documents, or important judgments. 50 credits per request.
The feature also changes consumption
Along with the model, which feature you use changes credit consumption. Agent mode and deep research ask AI to investigate, make decisions, and carry out work, so one request can use a significant number of credits.
Agent mode
30 credits per message
deep research
50 credits per task
Image generation
5 credits per image
Voice
5 credits per minute
Think of a taxi, a high-speed train, and a flight. All take you to a destination, but their fares differ. Agent mode and deep research are like asking AI to fly: investigate, decide, and report independently. The extra work means more credits.
Frequently asked questions
What are credits?
They are usage points used in ChatGPT business plans, deducted according to the features used.
Does consumption vary by model?
Yes. More capable models consume more. In the cited July 2026 table: Instant is unlimited, Thinking is 10 credits, and Pro is 50 credits.
Does it also vary by feature?
Yes. The cited table lists agent mode at 30 credits, deep research at 50, and images or voice at 5, with different billing units for each feature.
How much do light and demanding features differ?
One demanding request can cost the equivalent of 10–50 lighter operations. Choosing features for the task matters.
How can we save credits?
The first step is to stop using Pro or deep research for everything and choose according to the task.
Based on the official pricing pages of OpenAI and Google AI, as of June 2026. Prices, model names, and specifications can change; check current official information before deciding. Estimates use USD 1 = JPY 160 and approximately 1.5 tokens per Japanese character. They are illustrative Japan-market references, not quotations or estimates for English-language text. This series does not recommend any particular service. Contact sales to confirm current product pricing, delivery, warranty, and support for your country.
Company, product, and service names mentioned are trademarks or registered trademarks of their respective owners.