12 Days of OpenAI Kicks Off with ChatGPT Pro, Sora Launch, and More
OpenAI has launched its festive “12 Days of OpenAI” event, where for 12 days of December, the company will launch new product innovations for the AI space. Days 1-3 have already introduced exciting updates, including the launch of ChatGPT Pro, OpenAI’s Reinforcement Fine-Tuning Research Program, and the emergence of Sora out of research preview.
Day 1 of OpenAI’s 12 Days event featured the introduction of ChatGPT Pro, a $200 monthly plan that delivers scaled, unlimited access to OpenAI’s leading models and tools—including OpenAI o1, o1-mini, GPT-4o, and Advanced Voice. This plan serves to equip researchers, engineers, and other personas with ChatGPT’s most intelligent models for more reliable responses and accelerated productivity.
Offering o1 pro mode—an iteration of o1 designed to utilize more compute—ChatGPT Pro delivers powerful AI capabilities for verticals that rely on accurate, comprehensive answers, such as data science, programming, and case law analysis.
Day 2 of 12 Days of OpenAI announced the expansion of OpenAI’s Reinforcement Fine-Tuning Research Program, a program for testing Reinforcement Fine-Tuning on domain-specific tasks. Reinforcement Fine-Tuning is a new model customization technique that allows developers to configure OpenAI models with a variety of high-quality tasks, documenting the model’s response with reference answers.
The Reinforcement Fine-Tuning Research Program serves to cultivate models with the ability to excel at specific sets of complex, domain-specific tasks, according to OpenAI. It helps reinforce the model’s reasoning strategy with the data shared from organizations, presenting similar problems to enhance the model’s overall accuracy relating to domain-specific tasks.
Institutes, universities, and enterprises—especially those that deal with narrow, intricate sets of tasks led by experts—are encouraged to apply to the Reinforcement Fine-Tuning Research Program.
Day 3 of 12 Days of OpenAI announced an exciting, highly anticipated innovation—the launch of Sora, moving the video generation model out of research preview. Capable of creating incredibly realistic videos from text prompts, Sora understands and simulates reality—which reflects a foundational step in the progress of enabling models to interact with the real world, according to OpenAI.
Introduced earlier this year, Sora has been iterated in a new version—Sora Turbo—with greater performance capacity. Sora enables users to produce videos with quality up to 1080p resolution, up to 20 sec long, in widescreen, vertical, or square aspect ratios. Users can bring their own assets to extend, remix, or blend, or create entirely new video content. It is available as a standalone product for ChatGPT Plus and Pro users.
Though exciting, OpenAI presents Sora with a note of caution: The currently deployed version of Sora comes with limitations, such as unrealistic physics or a tendency to struggle with complex actions over a long period of time. Its launch is part of an effort to explore its capacities, as well as co-develop norms and safeguards, according to OpenAI.
It’s also worth noting that OpenAI is taking into consideration the ethical implications of video generation technology. All videos created with Sora are embedded with C2PA metadata, which helps provide transparency regarding the video’s origin. Additionally, OpenAI is blocking various “damaging forms of abuse,” including “child sexual abuse materials and sexual deepfakes” with various monitoring tools. More information about OpenAI’s safeguards for Sora can be found in its system card.
For more information about OpenAI’s latest innovations, please visit https://openai.com/. To stay up-to-date with the 12 Days of OpenAI, please visit https://www.enterpriseaiworld.com/.