OpenAI o3 will be retired from ChatGPT on August 26, 2026. Until its retirement, the o3 credit rate listed below continues to apply. This change does not affect API access. For more information, see: Model Release Notes.
Overview
This article outlines the current credit rates for ChatGPT Business and Enterprise/Edu plans under flexible, credit–based pricing. It covers ChatGPT models and features, ChatGPT for Excel, ChatGPT for PowerPoint, Workspace Agents, ChatGPT Work, and Codex.
For more information about how credits and flexible pricing work, see Flexible pricing for the Enterprise, Edu and Business plans.
If your Enterprise agreement specifies usage-based billing in U.S. dollars instead of credits, use the rate card identified in your agreement or contact your OpenAI representative. For Plus or Pro credit rates, see Using Credits for Flexible Usage in ChatGPT (Free/Go/Plus/Pro).
How credit-based pricing works
Depending on the feature, usage is measured as one of the following:
A fixed number of credits per message, task, generation, or connected minute.
Credits per 1M input tokens, cached input tokens, and output tokens.
Tokens are small units of information that OpenAI models process and generate. For token-based credit rates, total credit usage is calculated as follows:
Total credits = (input tokens / 1,000,000 × input rate) + (cached input tokens / 1,000,000 × cached input rate) + (output tokens / 1,000,000 × output rate)
Actual credit usage varies based on the model, task complexity, input size, cached input, output length, automations, and fast mode. To learn more, see What are tokens and how to count them?
Chat models
The credit rate for a ChatGPT message is based on the model used by the selected option.
| Model | Unit | Credits (approximately) |
|---|---|---|
| Instant | 1 message | N/A — Unlimited |
| GPT-5.6 Luna | 1 message | N/A — Unlimited |
| GPT-5.6 Sol | 1 message | 10 credits |
| GPT-5.6 Sol Pro | 1 message | 50 credits |
Instant chat is unlimited. Medium, High, and Extra High use different reasoning efforts, but they all use GPT-5.6 Sol and have the same rate of 10 credits per message. Selecting a higher reasoning effort does not increase the per-message credit rate. Pro uses GPT-5.6 Sol Pro and is charged at 50 credits per message.
When automatic reasoning is enabled, ChatGPT may use additional reasoning for a request while Instant remains selected. Requests that use GPT-5.6 Sol consume 10 credits. Requests that use GPT-5.6 Luna do not consume credits.
Unlimited or virtually unlimited access is subject to abuse guardrails and any model-specific usage allowances. Temporary restrictions may apply to prevent misuse.
ChatGPT features
| Feature | Unit | Credits (approximately) |
|---|---|---|
| Agent mode | 1 message | 30 credits |
| Deep research | 1 task | 50 credits |
| Images | 1 generation | 5 credits |
| Voice in ChatGPT | 1 minute | 5 credits |
For ChatGPT Edu customers, the following usage is included at no additional cost and does not consume credits:
Images: 3 generations per rolling 24-hour period.
Deep research: 5 queries per rolling 24-hour period.
The 24-hour period begins at the time of first use. For example, if a user generates an image at 10:00 AM, they can generate two additional images until 10:00 AM the following day. Usage beyond these limits consumes credits at the rates listed above.
ChatGPT Work and Codex
ChatGPT Work and Codex usage is priced based on token usage, calculated as credits per 1M input tokens, cached input tokens, and output tokens.
This rate card applies to:
ChatGPT Business users on standard ChatGPT seats, and users on Codex-only seats in eligible Business workspaces.
New and existing Enterprise, Edu, Health, Gov, and ChatGPT for Teachers customers.
A small subset of Enterprise customers that should continue using the legacy Codex rate card until their workspace is migrated. If you are unsure which rate card applies, contact OpenAI Sales.
For ChatGPT Business, standard ChatGPT seats use included plan limits first. If the workspace has purchased credits, eligible usage can continue from the shared credit pool after an included limit is reached. Usage-based Codex-only seats include Codex usage only and require workspace credits from first use.
Enterprise and Edu workspaces on flexible pricing scale usage with credits and do not have fixed rate limits; workspaces without flexible pricing remain subject to their plan limits.
Reasoning choices such as Ultra are not separate model rows in this rate card: Ultra uses maximum reasoning and may run additional agents for eligible users, so credit usage still depends on the model used and the tokens produced by the task and any agents it runs.
Upcoming model retirement: On Aug 31, 2026, GPT-5.4 and GPT-5.4 mini will retire in Codex for users signed in with ChatGPT. Use GPT-5.6 Terra instead of GPT-5.4 and GPT-5.6 Luna instead of GPT-5.4 mini. OpenAI API access and Codex use with your own API key are not affected.
| Model | Input tokens (credits per 1M) | Cached input tokens (credits per 1M) | Output tokens (credits per 1M) |
|---|---|---|---|
| GPT-5.6 Sol | 125 credits | 12.50 credits | 750 credits |
| GPT-5.6 Terra | 50 credits | 5 credits | 300 credits |
| GPT-5.6 Luna | 5 credits | 0.5 credits | 30 credits |
| GPT-5.5 | 125 credits | 12.50 credits | 750 credits |
| Daybreak Blue | 125 credits | 12.50 credits | 750 credits |
| Daybreak Red | 312.5 credits | 31.25 credits | 1,875 credits |
| GPT-5.4 | 62.50 credits | 6.250 credits | 375 credits |
| GPT-5.4 mini | 18.75 credits | 1.875 credits | 113 credits |
| GPT-5.3-Codex | 43.75 credits | 4.375 credits | 350 credits |
| GPT-5.2 | 43.75 credits | 4.375 credits | 350 credits |
| GPT-Image-2 (image) | 200 credits | 50 credits | 750 credits |
| GPT-Image-2 (text) | 125 credits | 31.25 credits | 250 credits |
These rates apply to supported ChatGPT Work and Codex activity, including local tasks, cloud tasks, automations, code review, auto review, and delegated workers. Charges are based on the model used and the actual input, cached input, and output tokens consumed.
Notes
Codex does not charge for cache writes.
Daybreak access requires approval through the OpenAI Daybreak/Trusted Access for Cyber program. Daybreak Blue uses the same rates as GPT-5.6 Sol. Daybreak Red requires separate approval and provisioning.
A typical Codex task using GPT-5.6 Sol may consume between 5 and 40 credits per task.
Fast mode consumes credits at a higher rate for supported models. See Speed for current rates.
Code review uses GPT-5.3-Codex.
Auto review uses GPT-5.4.
GPT-5.3-Codex-Spark is available to ChatGPT Pro users only as a research preview, uses a separate usage limit, and is not available through API keys at launch. For Plus and Pro credit rates, see Using Credits for Flexible Usage in ChatGPT (Free/Go/Plus/Pro).
Codex, ChatGPT Work, ChatGPT for Excel, and Workspace Agents draw from the same agentic usage and credit pool when those features are available on your plan. See Codex pricing and usage limits.
On average, Codex costs approximately $100–$200 per developer per month, though actual costs vary significantly depending on the model used, the number of concurrent instances, automations, and fast mode. See best practices for maximizing rate limits and managing token consumption.
ChatGPT for Excel, PowerPoint, and Workspace Agents
Note: Workspace Agent, ChatGPT for Excel/Sheets, and ChatGPT for PowerPoint use token-based pricing. The rates below apply to Business and Enterprise plans.
Tasks and runs do not have a fixed credit cost. Credit usage depends on the mix of input tokens, cached input tokens, and output tokens used. The final amount varies by task complexity, input size, cached input, and output length.
Example calculation
If a GPT-5.5 Workspace Agent run uses 20,000 input tokens, 80,000 cached input tokens, and 5,000 output tokens, the run consumes about 7.25 credits:
Input: 20,000 / 1,000,000 × 125 = 2.5 credits.
Cached input: 80,000 / 1,000,000 × 12.50 = 1 credit.
Output: 5,000 / 1,000,000 × 750 = 3.75 credits.
Total: 7.25 credits.
ChatGPT for Excel/Sheets
| Model | Input tokens (credits per 1M) | Cached input tokens (credits per 1M) | Output tokens (credits per 1M) |
|---|---|---|---|
| GPT-5.5 | 125 credits | 12.50 credits | 750 credits |
A typical ChatGPT for Excel/Sheets task using GPT-5.5 may consume between 5 and 20 credits per message.
ChatGPT for PowerPoint
| Model | Input tokens (credits per 1M) | Cached input tokens (credits per 1M) | Output tokens (credits per 1M) |
|---|---|---|---|
| GPT-5.5 | 125 credits | 12.50 credits | 750 credits |
A typical ChatGPT for PowerPoint task using GPT-5.5 may consume between 10 and 50 credits per message.
ChatGPT Workspace Agents
| Model | Input tokens (credits per 1M) | Cached input tokens (credits per 1M) | Output tokens (credits per 1M) |
|---|---|---|---|
| GPT-5.5 | 125 credits | 12.50 credits | 750 credits |
| GPT-5.4 | 62.50 credits | 6.250 credits | 375 credits |
A typical end-to-end Workspace Agent run using GPT-5.5 may consume between 5 and 25 credits.
Included rate limits for ChatGPT Business
Standard ChatGPT Business seats include access to ChatGPT and Codex. Usage for ChatGPT Work, Workspace Agents, ChatGPT for Excel, ChatGPT for PowerPoint, and Codex draws from the same agentic usage and credit pool. Flexible pricing can be used to extend included limits once pricing is in effect for each feature. Usage-based Codex-only seats include Codex only. Learn more about Business seat types, including Codex seats, in What is ChatGPT Business?
ChatGPT Work follows the same usage structure as Codex. Because Codex examples are based on coding tasks, ChatGPT Work usage varies by task.
ChatGPT Sites
ChatGPT Sites is available to Business and Enterprise customers, with public publishing available where supported. Sites is also rolling out in public beta to additional paid plans. Pricing information will be available soon.
ChatGPT Voice in Desktop
ChatGPT Voice in Desktop uses a separate connected-minute meter. For ChatGPT Business, Edu, and Enterprise customers using credits or pay-as-you-go billing, connected Voice time is expected to cost approximately 6 credits per minute.
When Voice starts a Codex task, the task also uses the shared Codex usage and credit pool at the token-based rates above. Voice time and Codex task work are metered separately.
See ChatGPT Voice in Desktop pricing and limits for current plan allowances and additional details.
Monitoring usage and managing credits
You can monitor Codex & Work usage limits in Codex settings under Usage. Depending on your plan and workspace role, you may also be able to view remaining credits, purchase credits, or manage auto-reload there. If you cannot add credits yourself, ask your workspace owner or admin.
Actual credit usage varies based on the model used, task size, input and output mix, automations, fast mode, and the number of concurrent Codex instances.
Legacy rates
Legacy ChatGPT models
The following rates apply to legacy ChatGPT models and are billed in credits per message.
| Model | Unit | Credits (approximately) |
|---|---|---|
| o3 | 1 message | 10 credits |
| o3-pro | 1 message | 50 credits |
| GPT-5.3 | 1 message | N/A — Unlimited |
| GPT-5.4 Thinking mini | 1 message | N/A — Unlimited |
| GPT-5.3 Pro | 1 message | 50 credits |
| GPT-5.3 Thinking | 1 message | 10 credits |
| GPT-5.4 Thinking | 1 message | 10 credits |
| GPT-5.4 Pro | 1 message | 50 credits |
| GPT-5.5 Instant | 1 message | N/A — Unlimited |
| GPT-5.5 Thinking | 1 message | 10 credits |
One prompt and response form a single message. Unlimited or virtually unlimited access is subject to abuse guardrails and any model-specific usage allowances. Temporary restrictions may apply to prevent misuse.
Legacy Codex rate card
Most new and existing customers across all plans have been migrated to token-based credit pricing and should use the ChatGPT Work and Codex rate card above. A small subset of Enterprise customers should continue using this legacy rate card until their workspace is migrated. For more information, contact OpenAI Sales.
The legacy rate card expresses Codex usage as approximate average credits per message or pull request. Actual credit usage can vary based on task size, model choice, and reasoning requirements.
| Activity | Unit | GPT-5.6 Sol | GPT-5.6 Terra | GPT-5.6 Luna | GPT-5.5 Cyber | GPT-5.5 | GPT-5.4 |
|---|---|---|---|---|---|---|---|
| Local tasks | 1 message | ~14 credits | ~6 credits | ~1 credits | ~56 credits | ~14 credits | ~7 credits |
| Cloud tasks | 1 message | Not available | Not available | Not available | Not available | Not available | ~34 credits |
| Code review | 1 pull request | Not available | Not available | Not available | Not available | Not available | ~34 credits |
These averages also apply to legacy GPT-5.2, GPT-5.2-Codex, GPT-5.1, GPT-5.1-Codex-Max, GPT-5, GPT-5-Codex, and GPT-5-Codex-Mini.
FAQ
Which rate card should I use?
Use the rates in this article if your workspace uses credit-based flexible pricing. If your Enterprise agreement specifies usage-based billing in U.S. dollars, use the rate card identified in your agreement, or refer to this Help Center article. A small subset of Enterprise customers should continue using the legacy Codex rate card until their workspace is migrated. If you are unsure, contact your OpenAI representative or OpenAI Sales.
Why are some rates fixed while others are token-based?
Some ChatGPT experiences are billed as a fixed number of credits per message, task, generation, or minute. Agentic products and Codex are billed by the input, cached input, and output tokens used, which provides a direct mapping between model activity and credit consumption.
Does selecting a higher ChatGPT reasoning effort increase the message rate?
No. Medium, High, and Extra High use different reasoning efforts, but they use GPT-5.6 Sol and have the same 10-credit message rate.
How does token-based pricing affect my credit usage?
The impact depends on your workload mix. Some users may see higher credit consumption and others may see lower consumption depending on input, cached input, and output usage. Output-heavy tasks and fast mode generally consume more credits than lighter tasks.