Learn how token rates work for ChatGPT models and features across Chat, ChatGPT Work, and Codex. This rate card covers model and feature pricing across Chat, Work, and Codex. For the billing model and comparison with credit-based agreements, see: Token-based billing for ChatGPT Enterprise . Customers on credit-based pricing plans should refer to ChatGPT Rate Card (Credit based Pricing) . For questions about your plan, please reach out to your OpenAI representative. On these Enterprise agreements, Instant usage is billable at the applicable rates, subject to any discounts in your agreement. What are tokens? Tokens are small units of information that OpenAI models process and generate. Token-based pricing measures usage across input tokens, cached input tokens, and output tokens. Your total charge is calculated by applying the applicable rate for the model and feature used to each token category, then adding those amounts together. You can learn more about how tokens are calculated here . Prices in USD per 1M tokens. Prices are expressed per million input, cached input, and output tokens. These rates replace ChatGPT message rates which were previously based on model and message count . Messages sent in chat will be based on token usage and model, so that usage pricing is standardized across all product surfaces. Prices are expressed per million input, cached input, and output tokens. These rates replace ChatGPT message rates which were previously based on model and message count . Messages sent in chat will be based on token usage and model, so that usage pricing is standardized across all product surfaces. The total cost of a request is calculated as follows: cost = (input tokens / 1,000,000 × input rate) + (cached-input tokens / 1,000,000 × cached-input rate) + (output tokens / 1,000,000 × output rate) The total cost of a request is calculated as follows: cost = (input tokens / 1,000,000 × input rate) + (cached-input tokens / 1,000,000 × cached-input rate) + (output tokens / 1,000,000 × output rate) GPT-5.6 rates shown in the table include reasoning efforts Instant , Medium , High , and Extra High and Pro. Selecting a higher reasoning effort does not increase the rate. GPT-5.6 rates shown in the table include reasoning efforts Instant , Medium , High , and Extra High and Pro. Selecting a higher reasoning effort does not increase the rate. GPT-Rosalind-Research is only available to select Enterprise plans. Read more. GPT-Rosalind-Research is only available to select Enterprise plans. Read more. GPT-6.1 Sol, GPT-6 Sol, and GPT-6 Luna are models for ChatGPT Work and Codex. They are not available in Chat. Note: GPT-5.6 Sol’s promotional pricing is available at least through Nov 21, 2026. The promotion applies to token-based usage billed in USD and does not apply to legacy metering. Included plan usage, 5-hour and weekly limits, and legacy credit rates are unchanged. Prices in USD per 1M tokens. The rates below are for Standard mode. 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. Image generation, Voice, web search, long context, fast mode, and regional processing may create separate or additional charges as described below. * Access to Astra for Law is limited to selected law firms in the United States through Trusted Access . Prices apply for all reasoning levels available for a given model (light, medium, high, extra high, ultra) unless otherwise noted. Prices apply for all reasoning levels available for a given model (light, medium, high, extra high, ultra) unless otherwise noted. Calculate the token charge using the rates per 1 million tokens: Calculate the token charge using the rates per 1 million tokens: Token charge = (input tokens / 1,000,000 × input rate) + (cached input tokens / 1,000,000 × cached-input rate) + (output tokens / 1,000,000 × output rate) Other applicable feature charges and fees are added as described below. Daybreak access requires approval through the OpenAI Daybreak/Trusted Access for Cyber program . Daybreak Blue usage with GPT-5.6 Sol uses the GPT-5.6 Sol rates. Daybreak Red requires separate approval and provisioning. Daybreak access requires approval through the OpenAI Daybreak/Trusted Access for Cyber program . Daybreak Blue usage with GPT-5.6 Sol uses the GPT-5.6 Sol rates. Daybreak Red requires separate approval and provisioning. Code review uses GPT-5.3-Codex.

Code review uses GPT-5.3-Codex. Auto review uses GPT-5.6 Luna. Auto review uses GPT-5.6 Luna. GPT-Rosalind-Research is only available to select Enterprise plans. Read more. GPT-Rosalind-Research is only available to select Enterprise plans. Read more. Feature availability and applicability The feature rates below apply wherever the listed feature is available across Chat, Work, and Codex, unless a section states otherwise. A single request may include both model-token charges and separate feature charges. Prices in USD per 1M tokens. These image-generation rates apply wherever GPT-Image-2.0 is available across ChatGPT, Work, and Codex. Prices in USD per 1M tokens. Image generation usage is measured in tokens, with total cost based on input text tokens, input image tokens when editing an image, and image output tokens. Image output-token usage varies with the requested dimensions and quality, while edits that include reference images may require additional input tokens. Larger or higher-quality images generally use more tokens, although token usage can vary by resolution. For image generation cost estimates, use the calculator in the image generation guide. Connected minutes and model activity are metered separately. A Voice session may therefore incur both the connected-minute charge and token charges for the backend model and any delegated workers. ChatGPT for Word, Excel/Sheets, and PowerPoint These features are billed at API model rates for the input, cached input, and output tokens used. Rates apply to models supported by each feature. From Sep 17, 2026 through Sep 30, 2026, ChatGPT Enterprise customers can use GPT-5.6 Sol in ChatGPT for Word during a free preview. This preview applies only to Word. Standard rates, in USD per 1 million tokens: GPT-5.6 Sol is the default model where available. Model availability depends on your region and workspace settings. Workspace admins can enable or disable access to each model. Codex, ChatGPT Work, Word, Excel, PowerPoint, and Workspace Agents use the applicable model and feature rates in this article and share agentic usage where available on your plan. ChatGPT Workspace Agents Prices in USD per 1M tokens. Usage across Codex, ChatGPT Work, ChatGPT for Excel, ChatGPT for PowerPoint, and Workspace Agents is charged using the applicable model and feature rates in this article. Note that these products share the same agentic usage when they are available on your plan. The following fees and multipliers may apply in addition to the standard model rate. GPT-5.6 and GPT-5.5 support long context in Work and Codex. Long context is not available in the Chat tab. GPT-5.6 and GPT-5.5 support long context in Work and Codex. Long context is not available in the Chat tab. GPT-6 Astra — Codex long-context exception GPT-6 Astra usage in Codex does not incur additional long-context multipliers above 272K input tokens. Codex does not charge for cache writes. You can monitor your Codex usage in the global admin console . For information about rate limits and reducing token consumption, see Codex pricing and usage limits and best practices for managing token consumption . Actual costs vary based on the model used, task size, input and output mix, automations, fast mode, and the number of concurrent Codex instances. For workspace budgets and user limits, see: Manage usage limits and overages in ChatGPT Enterprise and Edu . Was this article helpful?