OpenAI, the company behind ChatGPT, has unveiled a series of pre-holiday events, promising new AI features and products every day over 12 days. This exciting initiative kicked off on December 5 at 10 a.m. Pacific Time (8 p.m. Kyiv time).
The company teased a mix of significant launches and smaller, intriguing updates. OpenAI CEO Sam Altman shared on X (formerly Twitter):
“Every day we will be live-streaming new products or demos — including both large-scale launches and nice little things. We have a lot to show, we hope you enjoy! Merry Christmas.”
While OpenAI has kept most details under wraps, the community anticipates major releases, such as the text-to-video tool Sora and advancements in logical reasoning models. Smaller updates might include a festive Santa-style ChatGPT voice and holiday-themed features.
Day One: A Strong Start
OpenAI opened its advent calendar with a major update to the o1 logical thinking model, codenamed Strawberry, and the introduction of a new ChatGPT Pro subscription plan.
The o1 model, now widely available, is faster, more powerful, and more accurate than its preview version. It excels in coding, mathematics, and image processing. Users of ChatGPT Plus and Team tariffs can access this update immediately, with availability expanding to Enterprise and Edu customers next week.
A new ChatGPT Pro plan was also introduced for $200 per month, offering unlimited access to o1, GPT-4o, Advanced Voice mode, and an exclusive pro version of o1 optimized for solving complex problems. Meanwhile, the $20-per-month Plus plan remains available, providing early access to new features but excluding o1’s pro mode.
Additionally, OpenAI announced the ChatGPT Pro Grant Program to support medical research, with plans to expand into other fields.
Day Two: Reinforcement Fine-Tuning
On the second day, OpenAI expanded its reinforcement fine-tuning research program. This initiative allows developers and researchers to train AI models for specialized tasks in specific industries.
What Is Reinforcement Fine-Tuning?
This method involves setting specific tasks and providing feedback to refine the AI’s behavior. For example, training an AI to analyze legal documents involves providing examples and evaluating the model’s responses. Over time, the model improves its accuracy and generalizes better for similar tasks.
This approach is ideal for fields like medicine, finance, or scientific research. OpenAI has opened the program to research institutes, universities, and enterprises working on well-defined, narrow tasks that could benefit from tailored AI solutions.
Day Three: Text-to-Video Tool Sora
The third day marked one of the most anticipated releases: Sora, OpenAI’s text-to-video tool. Initially showcased in February, Sora is now widely available.
New Features in Sora:
- Storyboard: Enables users to create multiple videos and combine them into a single clip using an interactive timeline.
- Creativity Feed: A gallery where users can share videos, view others’ creations, and modify them based on shared queries.
Sora supports video style customization, looping, trimming, and multiple formats, including vertical, horizontal, and square. Subscription users can access video generation:
- ChatGPT Plus ($20/month): 50 videos/month, up to 5 seconds each, 720p resolution.
- ChatGPT Pro ($200/month): 500 videos/month, up to 20 seconds each, 1080p resolution, with up to five variations per query.
Non-subscribers can browse the Creativity Feed but cannot create videos, notes NIX Solutions. To use Sora, visit Sora.com and log in with a ChatGPT Plus or Pro account.
Day Four: Canvas Feature Expansion
On day four, OpenAI introduced enhancements to Canvas, a feature launched in October. Designed for programming and writing tasks, Canvas now integrates directly into ChatGPT, opening in a separate window.
Canvas acts as an intelligent assistant, providing in-context recommendations and comments for projects. It can be accessed via the ChatGPT text box or manually activated with a chat request.
Previously limited to premium users, Canvas is now available to everyone, including free-tier users, through the web version and the ChatGPT Windows app. macOS compatibility is expected soon.
The 12-day event continues, promising exciting announcements. Stay tuned—we’ll keep you updated as more integrations and features are revealed!