OpenAI’s 12 Days of Announcements: What to Expect

Updated 5 November 2026 · Standard formula, shown below

“12 Days of OpenAI” was OpenAI’s December 2024 announcement series: the company published a new product, feature, or update on each of 12 weekdays, from December 5 through December 20. It was a real, completed event—not a promise of 12 future announcements—and the releases ranged from major models to smaller tools and research previews.

What “12 Days of OpenAI” meant

OpenAI used the “12 Days of OpenAI” name for a run of livestreams and announcements during the 2024 holiday season. The schedule followed weekdays rather than 12 consecutive calendar days. The company introduced one main announcement on each day, sometimes accompanied by additional details or access information.

The phrase “new AI feature every day” is a useful shorthand, but it is not quite precise. Some days brought a new model or a substantial product capability; others focused on availability, a research preview, or a tool aimed at developers. Not every announcement was a feature that every ChatGPT user could immediately try.

The announcement schedule and what was revealed

OpenAI’s daily announcements covered a broad mix of products and research. The principal reveals were:

December 5: The o1 reasoning model became generally available in ChatGPT, alongside other updates including ChatGPT Pro.
December 6: OpenAI introduced its text-to-video model, Sora, to eligible users in supported regions.
December 9: The company announced reinforcement fine-tuning as a research preview for developers.
December 10: ChatGPT gained a video and screen-sharing experience for Advanced Voice, initially rolling out to eligible subscribers.
December 11: OpenAI announced ChatGPT integration with Apple Intelligence, with access dependent on Apple’s rollout and compatible devices.
December 12: The company introduced Projects in ChatGPT, a way to group related chats and files around ongoing work.
December 13: OpenAI released ChatGPT search more broadly and announced that it was available without signing in.
December 16: The company introduced Canvas, a workspace for working with writing and code, to more users and expanded access beyond its earlier testing.
December 17: OpenAI announced ChatGPT access through 1-800-CHATGPT in the United States, with a limited monthly call allowance, and messaging access through WhatsApp.
December 18: The company announced that its advanced voice mode could use Santa’s voice for a seasonal experience.
December 19: OpenAI introduced a video- and image-capable version of its Sora Turbo model through Sora.com for eligible users.
December 20: The final day featured OpenAI’s o3 and o3-mini reasoning models as research previews, rather than an immediate general release.

That list describes the headline announcements, not every related change in access or product detail. Availability differed by country, account type, platform, and rollout stage. For example, an announcement of a model did not necessarily mean that all users could select it that day.

Which announcements mattered most to users?

The most relevant release depended on what someone wanted to do. ChatGPT search addressed a common need: getting current web information with linked sources inside a conversational interface. Projects and Canvas focused on organization and sustained work, while Advanced Voice’s video and screen sharing connected conversation with what a user could show the app.

Sora was a major announcement for people interested in generating video from prompts or images, but its access was limited and subject to product rules and regional availability. The o1 and o3 announcements drew attention to reasoning models designed for more complex tasks. In o3’s case, the December announcement was a research preview; it should not be confused with the model’s later availability.

Other announcements were narrower. Reinforcement fine-tuning was aimed at developers exploring model adaptation, and Apple Intelligence integration depended on Apple’s supported software and device rollout. The phone and WhatsApp options offered another route to interact with ChatGPT, but did not replace the full range of app features.

How access and rollouts worked

OpenAI did not make every item available to everyone at the same time. Some capabilities were tied to paid plans, some were limited to particular regions or platforms, and some were labeled research previews. A feature could also roll out gradually after its announcement, so two users checking on the same day might see different options.

To check whether a specific announcement applies to you, look at the current product documentation or release notes and verify:

Whether the feature is a full release, a preview, or only an announced plan.
Which plan, country, device, or app version is supported.
Whether access is automatic or requires choosing a model, enabling a setting, or joining a preview.
What limitations apply to usage, content, uploads, or integrations.

These distinctions are especially important for announcements involving external products. The Apple Intelligence announcement, for instance, did not mean that every Apple device immediately gained ChatGPT features; compatibility and Apple’s own software availability mattered.

What the event did—and did not—signal

The event showed how broad OpenAI’s product work had become: language models, web search, image and video generation, voice, developer tools, and ways to organize chats appeared in the same series. It also illustrated why a daily announcement format can blur differences between a finished feature, a limited rollout, and a forward-looking research preview.

It did not establish a recurring annual event, guarantee that OpenAI would announce a feature every day in future Decembers, or mean that all announced technology was production-ready for every use. The December 2024 schedule is best understood as a one-time, clearly dated campaign. Anyone referring to “12 days” now should identify it as the 2024 series unless discussing a newly confirmed event.

Frequently Asked Questions

Did OpenAI really announce something every day for 12 days?
Yes. OpenAI held its “12 Days of OpenAI” series on 12 weekdays from December 5 to December 20, 2024, with a principal announcement on each day.

Are the 12 days happening now?
The series described here took place in December 2024. It is not an ongoing event or a standing guarantee of daily announcements. Check OpenAI’s official announcements for confirmation of any later event.

What was the biggest announcement?
There is no single objective answer. Sora, the o1 model’s general availability, ChatGPT search, and the o3 research preview were among the most prominent, but their importance depends on a user’s needs.

Could everyone use each new feature right away?
No. Access varied by plan, region, platform, and rollout timing. Some announcements were previews or integrations that required compatible devices or separate availability.

Did OpenAI release a new model on all 12 days?
No. The schedule included models, product features, integrations, access changes, and a seasonal voice experience. It was not 12 consecutive model launches.

Method: the standard formula for this calculation, rounded as each result states. See how the calculators work or report an error.