OpenAI's "12 Days of OpenAI" event showcased a range of new features and models, enhancing their AI offerings across various domains. Here's a summary of the key launches and their potential applications:
o1 Model Release
The full version of the o1 reasoning model was made available to ChatGPT Plus and Pro subscribers. This model offers improved performance across various prompts, including math and science, with fewer mistakes and faster processing. It also introduced multi-modal reasoning capabilities, allowing users to input and analyze images alongside text.
ChatGPT Pro Subscription
A new tier of ChatGPT subscription was introduced at $200 per month, offering unlimited access to OpenAI's best models and features, including o1-mini, GPT-4o, and Advanced Mode. This tier is designed for power users who require extensive AI capabilities for complex tasks.
Reinforcement Fine-Tuning Research Program Expansion
OpenAI expanded access to its Reinforcement Fine-Tuning program, allowing developers and organizations to customize models for specific, complex tasks. This feature is particularly useful for research institutes, universities, and enterprises working on specialized AI applications.
Sora Video Generator
The long-awaited Sora video generation model was officially released, capable of producing 20-second clips at 1080p resolution. While the initial reception was mixed, this tool opens up possibilities for content creators and marketers to generate short-form video content.
Canvas Improvements
The Canvas feature was integrated directly into the GPT-4o model and made available to users across all tiers. It now supports running Python code within the Canvas space and allows for the construction of custom GPTs, enhancing its utility for developers and data analysts.
Apple Intelligence Integration
ChatGPT was integrated into Apple Intelligence, specifically Siri, allowing iOS users to access the chatbot directly through their devices. This integration brings advanced AI capabilities to Apple's ecosystem, potentially enhancing the user experience for iOS users.
Advanced Voice Mode Enhancements
The conversational chatbot model gained the ability to view the world through a mobile device's camera or via screen sharing. This feature enables users to ask questions about their surroundings without manual description or photo uploads. Additionally, a seasonal Santa Claus voice was added to Advanced Voice Mode.
Projects for ChatGPT
A new smart folder system called "Projects" was introduced, allowing users to better organize their chat histories and uploaded documents by subject. This feature improves workflow management for users working on multiple AI-assisted tasks or projects.
ChatGPT Search for All Users
The ChatGPT Search function, which allows the AI to search the internet and synthesize information into conversational answers, was made available to all logged-in users regardless of subscription tier. This feature competes with similar offerings from other AI companies and enhances ChatGPT's utility as a general-purpose information tool.
API Access to o1 Model
Select developers were granted access to the full version of the o1 reasoning model through the API, along with real-time API updates and a new model customization technique called Preference Fine-Tuning. New SDKs for Go and Java were also released, expanding the development options for AI integration.
1-800-ChatGPT Phone Service
OpenAI launched a phone service allowing users to access ChatGPT's Advanced Voice Mode through a toll-free number, even from landlines or basic mobile phones. This service aims to make AI assistance accessible to those without internet access or smartphones.
Expanded Coding App Integration
ChatGPT's ability to interface directly with popular coding applications was significantly expanded, allowing it to pull code snippets from a wider range of apps and IDEs. This feature also extends to conventional text programs like Apple Notes and Notion, enhancing productivity for developers and writers alike.
o3 and o3-mini Models Preview
OpenAI teased its upcoming o3 and o3-mini reasoning models, which reportedly offer superior performance on challenging math, science, and coding benchmark tests compared to their predecessors. These models are currently in safety testing and red teaming trials, with no specific release date announced for general users.