Sharing feedback, evaluation and fine-tuning data, and API inputs and outputs with OpenAI

OpenAI aims to create models that are useful for developers across a variety of domains and applications. We improve our models through research advancements and by applying them to real-world challenges and contexts.

One way to contribute to this improvement is by opting to share feedback, evaluation and fine tuning data, or API inputs and outputs with OpenAI. Organizations who choose to do so can manage these options at the project or organization level through their data controls settings. Data you share will be used to help identify usage patterns, measure model quality, and inform future evaluation and training of models.

You’re always in control of your data and privacy when using the OpenAI API. You can change your settings and choose to opt out again at any time. To help protect people’s privacy, we take steps to reduce the amount of personal information in our training datasets before they are used for any model improvement and training.

Please be aware that by sharing data from your organization to OpenAI, you confirm that you have the appropriate permissions for OpenAI to process and use this data as described in this article. Please do not include any sensitive, confidential, or proprietary information in the data you share.

You will need to be an organization owner to opt in to share data with OpenAI. Accounts with Zero Data Retention enabled cannot opt in to our data sharing mechanisms.

The ability to share feedback via the playground is disabled by default for all organizations. If an account owner wants to opt-in to sharing API model feedback for their organization, they can select “enabled” or “enabled for selected projects” on the organization settings page.

When this is set to “enabled,” members of an organization will see a “thumbs down” button on model responses in the Playground. When clicked, this will give the user the option to share feedback and the conversation up to that point (including inputs, outputs, and files uploaded) with our feedback systems.

By default, data sharing for evaluation and fine-tuning data is disabled for all organizations. Account owners can opt into this feature via their organization’s settings page, either for all projects or only selected ones. Account owners can change these settings at any time and choose to opt out again. This setting is not available to certain organizations, including customers with Zero Data Retention enabled.

When set to 'Enabled,' evaluations and fine-tuning data sent to OpenAI on enabled projects will be shared with OpenAI. Evaluations you share with OpenAI are currently processed at no cost for up to 7 runs per week. When sharing data for reinforcement fine-tuning jobs, the resulting models will be billed at discounted inference rates.

By default, data sharing inputs and outputs is disabled for all organizations. Account owners can opt into this feature via their organization’s settings page, either for all projects or only selected ones. Account owners can change these settings at any time and choose to opt out again. This setting is not available to certain organizations, including Enterprise and customers with Zero Data Retention enabled.

When set to 'Enabled,' inputs and outputs that are sent to OpenAI on enabled projects will be shared with OpenAI.

Some organizations may qualify for daily complimentary tokens on traffic shared with OpenAI.

You can see if you’re eligible for the offer by going to your data sharing settings page and confirming whether you see the “You're eligible for free daily usage on traffic shared with OpenAI” offer. When you enable data sharing and you’re qualified for complimentary tokens, you'll see “You’re enrolled for complimentary daily tokens”.

This is what the offer for eligible users in Tiers 3-5 looks like:

Similarly, this is what the offer for eligible users in Tiers 1-2 looks like:
​

If you do not see this language on your data sharing page, you are not eligible for the complimentary tokens at this time.

If you opt-in to share traffic with OpenAI and qualify for the complimentary tokens offer, the complimentary tokens will apply automatically to traffic you share with OpenAI on the models listed below.

Note that you'll only get the complimentary tokens on traffic shared with OpenAI, so if you only opt-in to share traffic on select projects, only usage on those projects would qualify for the free tokens. You need a positive account balance to use OpenAI models and the free tokens.

You can confirm your free token usage by comparing usage activity (where the free tokens will be included in your total) with Costs (where the free tokens will not appear as a cost) on your Usage Dashboard. Make sure to select both “input tokens” and “output tokens” to view your full usage on the Usage Dashboard.

You can also view the complimentary tokens on your https://platform.openai.com/usage/chat-completions dashboard and grouping by “service tier.” The free tokens will be displayed as “data sharing incentive tier - input/output tokens” Make sure to include both input and output tokens to see total usage. Please note that the total number of data sharing incentive tier tokens may not be exactly 1M or 10M depending on request size (see “How does the token quota limit work and when does it refresh?” below).

This offer is only available on the following models, and the token quota is shared across the model groups. Fine-tuned models, fine-tuning training, evals, and tool use are not included.

You’re entitled to up to 1M (250k for tiers 1-2) tokens per day shared across gpt-4.5-preview, gpt-4.1, gpt-4o, o1, o3 and o1-preview models, and up to 10M (2.5M for tiers 1-2) tokens per day shared across gpt-4.1-mini, gpt-4.1-nano, gpt-4o-mini, o1-mini, o4-mini, o3-mini, and codex-mini-latest models.

If you use more than the combined limit of 1M/250k or 10M/2.5M across these model groups, overage will be billed at normal rates. For example, if you use 750k tokens on o1 and then 500k tokens on gpt-4o on one day, you’d be billed for 250k tokens as overage.

For example, for the 1M token daily quota:

In this example, the first seven requests are free. The eighth request would take the total past 1 million tokens, so it is charged in full. The free‑token counter resets at 00:00 UTC each day.

As of March 6, 2025, tokens refresh daily at 00:00 UTC time. We made this change to allow for improved predictability and observability.

You can opt out at any time on your data sharing settings page.

Only Org Owners have access to the setting and the ability to turn it on / off.

You can read more about how your data is handled in this Help Center article and on our Enterprise Privacy page.