- Sam Altman made a surprise announcement early morning on X.
- Starting December 5, the company behind ChatGPT is launching a campaign called 12 Days of OpenAI.
- It will feature daily live streams of new features or product announcements every weekday for two weeks.
OpenAI CEO Sam Altman confirmed the 12 days of announcements onstage at The New York Times’ DealBook conference on Wednesday morning. He didn’t say exactly what was coming though. OpenAI plans to launch or demo something every day for 12 days straight.
He shared the same message on X (formerly Twitter), “Starting tomorrow at 10 am Pacific, we are doing 12 days of OpenAI.”
What to Expect: Sora, o1 Reasoning Model, and More
While official details are scarce, there are strong indications of what might be unveiled:
Sora: OpenAI’s Text-to-Video Model
First up is Sora, OpenAI’s text-to-video model that can generate detailed, photorealistic videos from simple text prompts. Some alpha testers recently leaked its API access, protesting against what they called “art washing” of their creative work. There is a high chance for the service to be public.
o1 Reasoning Model: Full Version
Also on the list is the full version of the o1 reasoning model. The o1 model was announced back in September, however until now we only have access to the o1-Preview model but not the full version. The upcoming release could include live search data, improved memory integration, and enhanced project planning capabilities.
Holiday-Themed Advanced Voice Mode
Another expected release is a holiday-themed voice for ChatGPT’s Advanced Voice Mode. Users have spotted a snowflake icon in the code, hinting at a Santa-inspired twist just in time for Christmas.
“Operator” Autonomous AI Agent
OpenAI is set to launch “Operator,” an autonomous AI agent, at the start of 2025. Operator is designed to handle complex, multi-step tasks autonomously, such as coding, booking travel, and navigating web browsers. However, there’s also a good chance that this AI Agent may get released in the next 12 days.
GPT 4o-images
GPT-4o can generate images, but OpenAI currently locked it off. So we are still relying on DALL-E 3 which is outdated compared to models like Imagen 3 from Google. So we can expect OpenAI to finally unveil the model’s ability. However, all these are just expectations and speculations.
GPT-5 Rumors
Earlier speculation about a potential GPT-5 release has been dismissed by OpenAI, with the company denying reports of its arrival before the end of the year.
A Mix of Big and Small Gifts for AI Enthusiasts
Not every day will be a blockbuster. Leah Anise, an OpenAI spokesperson, mentioned that while some announcements will be significant—like Sora’s potential release—others will be smaller, designed to enhance existing tools like ChatGPT or Canvas. There’s also talk about Advanced Voice Vision, Canvas upgrades, and the availability of GPT-4o image generation, which until now has been inaccessible to most users.
So, what are your expectations from “12 Days of OpenAI”?
