ChatGPT made a surprise announcement that beginning on the December 5th, 2024, OpenAI would start unveiling and demonstrating its upcoming features to the public over 12 consecutive days. The announcement immediately attracted widespread attention, with the official launch of the highly anticipated o1 model being the most eagerly awaited highlight. Given the extensive range of products and features released, we have categorized and discussed them under the following five key areas.
Launch of New Models and Plans
On December 5th, OpenAI officially released the full version of the o1 model, replacing the previous o1-Previous version introduced in September. While the widely used GPT-4 series focuses on text comprehension and generation, o1 builds upon this by enhancing image analysis, processing, and reasoning capabilities. According to internal testing, the o1 model has reduced major errors by 34% compared to its preview version, while overall processing speed has increased by approximately 50%, demonstrating its superior performance. Additionally, OpenAI has launched o1 Pro, an advanced version of o1, which offers greater computational power for handling complex problems in specialized fields and deep learning models.
OpenAI has also introduced a new subscription plan—ChatGPT Pro. Previously, users could only choose between the Free and Plus plans. Under the Pro plan, users pay $200 per month for unlimited access to o1, GPT-4o, advanced voice features, and OpenAI’s most sophisticated model, o1 Pro, which delivers more precise and reliable responses across various domains. This plan is designed primarily for heavy users of ChatGPT, such as academic researchers and software developers, who require AI with high computational capacity. OpenAI has stated that it will continue to refine the Pro plan’s features.
Developer Optimizations
OpenAI has launched the "Enhanced Fine-Tuning Research Program", which is currently recruiting developers and machine learning engineers to train customized professional models. The goal is to leverage feedback from academic and industry experts to enhance the model’s reasoning capabilities in specific fields, improving accuracy and application depth. This initiative accelerates the integration of AI into industries such as finance, insurance, law, and engineering. OpenAI has already made significant progress in these areas, helping to boost research and analytical value.
In addition to model customization, OpenAI has optimized its developer tools, including:
- API Support for OpenAI o1 Reasoning Technology – Aimed at simplifying customer service, supply chain decision-making, and financial trend forecasting. This feature is intended for Level 5 users with expertise in deep learning applications and includes new functions such as function calling and structured output.
- Enhancements to Real-Time API – Providing a low-latency, natural conversational experience by integrating WebRTC technology while reducing costs.
- Preference Fine-Tuning – Allowing users and developers to adjust input and output data formats to increase flexibility.
- Expanded SDK Support – In addition to existing Python and Node.js options, OpenAI has introduced SDKs for Go and Java to broaden API development capabilities.
Innovations and Feature Enhancements
Sora
One of the most notable highlights of this release is Sora, a tool that automatically analyzes user prompts to generate videos tailored to their needs. Currently available only to ChatGPT Plus and Pro members, Sora introduces a groundbreaking feature—users can enter scene descriptions, and AI will generate corresponding videos (Prompt-Based Video Generation). Other editing features include Remix (video blending), Re-cut (re-editing), Loop (seamless looping), Blend (fusion effects), and Storyboard creation, most of which can be executed with simple text commands, making video editing more accessible than ever.
Sora has attracted significant attention for its breakthrough in extending generative AI capabilities to video content, driving AI closer to Artificial General Intelligence (AGI). This advancement may impact the demand for traditional video editing software, editors, and animators. However, there are widespread concerns regarding privacy, image rights, and copyright issues. OpenAI has stated that it employs a diverse set of training data sources and collaborates with experts to prevent misuse for illegal content creation. However, similar to the lawsuit OpenAI faced from The New York Times over unauthorized text usage for training ChatGPT, OpenAI has not disclosed the exact sources of Sora’s training data. Given these concerns, it may take time before Sora is fully adopted by industries and the general public.
Canvas
On the fourth day, OpenAI introduced Canvas, an editor-like interface available to all users. Traditionally, when writing articles or editing Python code, users had to repeatedly interact with ChatGPT, then manually copy and organize the responses in a separate editor. With Canvas, the left side of the interface serves as a ChatGPT chat panel, while the right side displays generated results. This setup allows users to enter prompts on the left while ChatGPT produces results on the right, eliminating the need for manual compilation and enhancing contextual accuracy.
Additionally, ChatGPT can refine results based on user requirements and adjust wording and formatting for different reading levels, from kindergarten to graduate students. However, existing professional editing tools like Notion already offer integrated AI assistance, and AI-powered coding tools like Cursor outperform Canvas in efficiency due to their complete code library context. Canvas is still in its early stages, and its competitive edge in practical applications remains to be seen.
New ChatGPT Features
1. Search Function
One of the most significant updates is the addition of a search function within ChatGPT. Users can now perform searches directly on the ChatGPT website, but the bigger breakthrough is the release of a Chrome extension that enables searches directly from the browser’s address bar. By partnering with major news and data providers, OpenAI ensures direct access to verified information, boosting credibility. This feature poses a direct challenge to Google’s dominance in the search engine market, where over 50% of revenue comes from search-driven advertising. A shift in user preference toward ChatGPT’s more intuitive search experience could disrupt Google’s business model.
2. 1-800 ChatGPT Hotline
U.S. users can now call the 1-800 ChatGPT hotline (1-800-242-8478), while global users can connect via WhatsApp. This feature surpasses internet limitations by providing AI assistance even in low-connectivity situations. However, security concerns remain, as similar past projects, like Google-411, were used to collect voice samples. Although OpenAI promises not to use these calls for large-scale language training, public skepticism persists, especially regarding privacy issues and potential conversation monitoring.
3. Project Management Function
OpenAI has introduced a project management feature that allows users to create organized folders for conversations, upload relevant documents, and build a customized AI knowledge base. ChatGPT can then extract key points from the uploaded data for analysis. This feature enhances contextual continuity in conversations and is comparable to Google’s Notebook LM, which also supports AI-assisted research and file uploads. Both tools are still in early development, and their effectiveness will depend on improvements in logical reasoning.
Partnership with Apple
On the 9th, OpenAI demonstrated how Apple Intelligence will integrate with ChatGPT, transforming it into an intelligent assistant across iOS and macOS systems. Similar to the features Apple showcased during its fall launch event, the key highlight is the expanded capabilities and depth of Siri’s applications. Unlike ChatGPT, which primarily relies on text input, Apple users can continue using Siri as they always have—issuing voice commands, which are then processed and analyzed by ChatGPT before Siri executes the requested action on the device.
For example, if a friend shares a message about a singer releasing a new album, the user simply needs to say "Play", and ChatGPT will understand the context and open the music app to play the album. This integration enhances both interactivity and execution speed, making interactions more intuitive and efficient. Another major feature is visual-based AI assistance—ChatGPT can analyze real-time input from the camera, providing users with relevant information based on what they see. This advancement further enhances Apple’s “Camera Shortcut” feature introduced in iPhone 16.
On the 11th, Apple followed up with the release of iOS 18.2 to support OpenAI’s new functionalities. Previously, the public remained skeptical about the real-world applications of Apple Intelligence, and even Apple’s Q3 earnings report had raised concerns among investors. However, the official launch of these features reassures the market and raises expectations for further improvements.
Despite the added convenience for Apple users, privacy concerns remain a key issue. As these AI-powered features analyze user habits and behavioral patterns in real time, they also raise questions about data security. Apple has long been known for its strong focus on privacy and security, and its partnership with ChatGPT may cause some users to question whether their personal data remains as protected as before.
Additional Features
On the eleventh day, OpenAI introduced a native desktop application for macOS and Windows, along with a new voice mode to simplify the input process and reduce barriers for users. OpenAI also expanded third-party integrations, covering multiple platforms, text editing apps, and Integrated Development Environments (IDEs). With these advancements, ChatGPT is evolving beyond just a Q&A tool—it functions like a secretary or assistant, working alongside users to provide proactive, real-time support.
Developers stand to benefit the most from these updates, as ChatGPT can now seamlessly integrate with coding environments, generating code based on user requirements and contextual understanding, performing debugging, and significantly improving coding efficiency.
In addition to the above features, this release also includes a limited-time "Santa Mode", allowing users to chat with Santa Claus on ChatGPT during the Christmas season.
On the final day of announcements, OpenAI reaffirmed its commitment to collaborating with legal and academic experts to further enhance AI's ability to recognize and process user inputs responsibly. This initiative aims to prevent misuse by malicious actors and strengthen overall safety measures.
Conclusion
Based on the above, the newly introduced ChatGPT models and applications are summarized in the following plans:
| Model/Application | Free | Plus | Pro | Team |
|---|---|---|---|---|
| GPT-4o | V (Limited) | V (Higher quota) | V | V (Higher quota) |
| o1-mini | V (Limited) | V | V (Limited) | |
| o1 | V (Limited) | V | V (Limited) | |
| o1 Pro | V | |||
| Sora | V (Limited) | V | ||
| Canvas | V | V | V | V |
| Search Function | V | V | V | V |
| Project Management Function | V | V | V |
The release of OpenAI’s latest products once again redefines the potential applications of AI. It demonstrates OpenAI’s vision not only to assist individuals in solving challenges in daily life and work but also to evolve AI into an independent-thinking assistant capable of exchanging ideas and learning alongside humans.
Currently, AI’s average intelligence is approaching human levels (with human IQ typically ranging between 90 and 110). As AI innovation accelerates, it will reshape the way humans access information. Highly intelligent AI could significantly replace labor demand in certain industries. Moreover, as people rely on AI for more assistance, they may need to disclose increasing amounts of personal data, heightening privacy risks. In the long run, this raises not only ethical and legal concerns but also fears that AI could eventually surpass human control.
As speculated by many, OpenAI officially announced on the 27th through its blog that it plans to restructure its for-profit division into a Delaware Public Benefit Corporation in 2025. This transition aims to prioritize the company’s mission and shareholder interests while maintaining the development of its nonprofit sector. OpenAI was originally established to create safe and widely beneficial AGI, and its choice of a nonprofit operational model was intended to ensure AI development remains within appropriate boundaries. While financial backing allows OpenAI to invest more in research, it also introduces profit-driven incentives. Therefore, as OpenAI shifts toward a for-profit model, the biggest challenge will be balancing business profitability with the advancement of technology for humanity—a matter that governments and the public must continue to monitor closely.
For more introduction about OpenAI, you can read more at News Feature: OpenAI’s Transition Towards Profitability
