OpenAI's 12 Days of Shipmas: Daily AI Innovations & Product Releases

OpenAI has kicked off an exciting holiday-themed campaign, 12 Days of "Shipmas," featuring 12 days of announcements, product launches, and demos. Each day, a new update will be unveiled, showcasing the company’s latest innovations.

Be sure to check back daily to see what OpenAI has in store—new releases will be added to this article as they're announced.

Day 1: OpenAI Unveils Full Version of o1 and ChatGPT Pro 

Day 2: OpenAI Launches Reinforcement Fine-Tuning Research Program

Day 3: OpenAI Introduces Sora: A Game-Changing Text-to-Video AI

Day 4: OpenAI Expands Canvas to All Users with New Features

Day 5: OpenAI Reveals ChatGPT in Apple Intelligence

Day 6: OpenAI introduces advanced voice with video & Santa mode

Day 7: ChatGPT Launches Projects: Organize and Customize Your Conversations

Day 8: ChatGPT Web Search With Citations

Day 9: Holiday treats for developers

Day 10: 1-800-CHATGPT

Day 11: Work With Apps

Day 12: o3 preview & call for safety researchers

Day 12: o3 preview & call for safety researchers

For some reason I cannot paste the details of our article that we wrote here, which I believe is due to our platform’s limitations on the length of articles.

Instead, please visit OpenAI’s blogs on this release:

Deliberative Alignment and Early Access for Safety Training.

Applications for early access open on December 20, 2024, and close on January 10, 2025. You can apply here.

Day 11: Work With Apps

OpenAI has introduced a series of exciting updates for its desktop apps as part of its “12 Days of Shipmas” event. On Day 11, the focus was on enhanced desktop functionality, enabling ChatGPT to work seamlessly with other apps on users’ computers.

These updates are designed to make ChatGPT more integrated into workflows, offering automation and advanced support for coding, writing, and more.

ChatGPT Desktop Apps: A Big Step Forward

OpenAI’s desktop apps, first launched for Mac and later for Windows, now include advanced features tailored to working alongside other apps. Key highlights include:

• App Integration: ChatGPT can now interact directly with apps on your desktop. With user permission, it can pull context from open applications to generate tailored responses.

• Automation: Automate workflows like coding, data visualization, or even creating holiday-themed charts without the need to switch apps.

• Enhanced User Control: Users retain full control over what ChatGPT can access, ensuring privacy and security.

Key Use Cases

OpenAI demonstrated several ways ChatGPT works with desktop apps to simplify complex tasks:

• Coding Assistance: ChatGPT can now assist with tasks directly within coding environments like Xcode, VS Code, and JetBrains IDEs. For example, users can request coding solutions or troubleshoot issues while ChatGPT seamlessly integrates with their chosen IDE.

• Data Visualization: By integrating with tools like Warp, ChatGPT can generate custom commands and create charts or visualizations based on user prompts, such as analyzing repository activity.

• Writing Support: Applications like Notion, Apple Notes, and Quip are now compatible with ChatGPT. Users can highlight text, ask for context-specific improvements, and even fact-check their work using the search functionality.

Advanced Features

A major addition is Advanced Voice Mode, allowing users to interact with ChatGPT using voice commands. OpenAI showcased how this feature can help users refine documents, brainstorm ideas, or adjust playlists—all through conversational voice input.

The desktop apps also leverage OpenAI’s o1 reasoning model for more complex coding and logic-based tasks, offering step-by-step explanations alongside solutions.

Availability

The new features are available starting today on the ChatGPT Mac desktop app. A Windows version is expected soon, ensuring broader accessibility for users across platforms.

What This Means

OpenAI’s desktop updates mark a shift toward greater integration and usability, allowing ChatGPT to do more than answer questions. By working directly with other apps, ChatGPT is becoming a more powerful assistant for developers, writers, and everyday users.

With the ability to streamline workflows and automate repetitive tasks, these features showcase the potential of AI to act as a true collaborator, further blurring the line between human and machine teamwork.

Day 10: 1-800-CHATGPT

As part of its “12 Days of Shipmas” initiative, OpenAI unveiled new ways to interact with ChatGPT, making the AI assistant more accessible than ever. Day 10 introduced the ability to call ChatGPT directly via a toll-free phone number, 1-800-CHATGPT, or engage with it through WhatsApp. These updates aim to reduce barriers to AI accessibility and connect more users worldwide to OpenAI’s technology.

Bringing ChatGPT to Phones

OpenAI has continuously evolved ChatGPT’s availability, first launching it on the web, then expanding to iOS and Android apps, and later adding desktop apps for macOS and Windows. The latest move brings ChatGPT to traditional telephones and WhatsApp, offering users another convenient way to access the AI assistant.

Phone Access

Users in the U.S. can now call ChatGPT by dialing 1-800-CHATGPT (1-800-242-8478). This voice-based interaction allows users to experience ChatGPT’s conversational AI even without a reliable internet connection.

For example, one user asked ChatGPT to identify a colorful dome-shaped house seen during a road trip. ChatGPT accurately described it as the "Flintstone House," a unique architectural landmark in California.

WhatsApp Integration

Globally, users can interact with ChatGPT on WhatsApp by adding it to their contacts. During a demo, the team used WhatsApp to get recipe suggestions, including options for vegan and meat-based meals, showcasing ChatGPT’s versatility.

Current Limitations and Future Plans

While the new features make ChatGPT accessible to a broader audience, there are some limitations. Advanced features, such as image-based conversations and search tools, are currently only available on the ChatGPT app and web platforms.

"For now, these features are just available on our mobile app and on our website," explained OpenAI during the demo.

However, OpenAI hinted that these capabilities might eventually come to phone and WhatsApp channels as they continue to improve and expand the platform. This phased rollout ensures the core experience remains robust and reliable while advanced features are refined for wider deployment.

Broadening Accessibility

OpenAI’s goal is to make artificial general intelligence beneficial and accessible to everyone. These new channels reflect this mission by enabling users with diverse devices and connectivity options to engage with ChatGPT.

“We care a lot about continuing to reduce barriers for more people to try out AI and see what it can do for them,” said Kevin Whe, OpenAI’s product lead.

The team behind the project noted that it began as a hackathon idea but quickly became a polished product designed to bring ChatGPT to more users.

What This Means

The introduction of phone and WhatsApp access for ChatGPT underscores OpenAI’s commitment to expanding the reach of AI. By reducing barriers like internet dependency or platform limitations, OpenAI is opening doors for millions of potential users worldwide.

These updates also signal a broader trend toward making conversational AI omnipresent and easily accessible. From smartphones to flip phones to even rotary phones, OpenAI’s push to bring ChatGPT to multiple channels positions the technology as a versatile assistant for everyday life.

As OpenAI continues to refine ChatGPT’s capabilities, its accessibility initiatives could serve as a blueprint for how AI can integrate seamlessly into users’ daily routines, regardless of their preferred technology.

Day 9: Holiday treats for developers

As part of its "12 Days of Shipmas" announcements, OpenAI has introduced a series of major upgrades for developers, including the new OpenAI o1 model (out of preview), improved Realtime API capabilities, a new powerful fine-tuning method, and official SDKs for Go and Java. These updates focus on enhancing performance, cost-efficiency, and customization, enabling developers to build faster, smarter AI-driven applications.

Introducing OpenAI o1: A Model for Complex Reasoning

The headline update is OpenAI o1, a model designed to handle multi-step tasks with improved accuracy and efficiency. OpenAI describes o1 as the successor to its o1-preview model, which has already been used to build advanced AI applications, from optimizing supply chain decisions to streamlining customer support, and forecasting complex financial trends.

Key features of OpenAI o1 include:

• Function Calling: Allows the model to interact with external data and APIs, making it ideal for tasks like retrieving live information or triggering system actions.

• Structured Outputs: Ensures responses are delivered in a specific, reliable format (e.g., JSON Schema), useful for integrating AI with other systems.

• Developer Messages: Allows developers to provide specific instructions or context for the model, such as setting the desired tone, style, or behavior for its responses.

• Vision Capabilities: Enables the model to analyze and reason over images, unlocking applications in fields like manufacturing, science, coding, and software development.

• Lower Latency: By using 60% fewer reasoning tokens than o1-preview, o1 delivers faster responses, making it more efficient for real-world use cases.

Developers now also have more control over how the model thinks using a new “reasoning_effort” API parameter, which allows adjustments to processing time for complex tasks.

Performance Benchmarks: Faster, More Capable AI

OpenAI’s latest version, o1-2024-12-17, which is an updated, post-trained version of the model that was released in ChatGPT two weeks ago, sets state-of-the-art benchmarks across multiple categories, including:

• General reasoning: Achieving 75.7% accuracy on GPQA diamond, up from 73.3%.

• Coding: Improving SWE-bench verified results from 41.3% to 48.9%, a significant gain.

• Mathematics: Achieving 96.4% accuracy on the MATH benchmark, compared to 85.5% previously.

• Vision: Adding strong performance in visual reasoning tasks like MathVista and MMMU.

• Factuality: Maintaining consistent performance with 42.6% accuracy on SimpleQA.

• Agents: Excelling in task-based benchmarks, with 73.5% on TAU-bench (retail) and 54.2% on TAU-bench (airline).

These benchmarks demonstrate o1’s ability to tackle increasingly complex, multi-step tasks while maintaining faster response times and higher efficiency.

Access is being rolled out gradually as OpenAI works to expand availability to more usage tiers and increase rate limits. To get started, visit the API documentation.

Realtime API Updates: Real-Time, Low-Latency Voice Interactions

OpenAI’s Realtime API has received major improvements to help developers build real-time, natural conversational experiences, such as voice assistants, live translation tools, customer support sytems, or virtual tutors. New updates include:

• WebRTC Support: WebRTC (Web Real-Time Communication) is an open standard that allows developers to build and scale low-latency voice applications with ease. It improves voice streaming quality, even under poor network conditions, making it ideal for mobile apps and IoT devices. It manages audio encoding, streaming, noise suppression, and congestion control seamlessly. With WebRTC, you can easily integrate Realtime capabilities using just a few lines of JavaScript

• Cost Reductions: Audio token pricing for GPT-4o has dropped by 60% to $40/1M input tokens, and cached audio costs are down 87.5% to $2.50/1M input tokens. This significantly reduces costs for developers building audio-based tools.

• GPT-4o Mini: A cost-efficient model that brings rich voice capabilities to the Realtime API at a fraction of the price while maintaining the same voice experiences. GPT-4o mini audio is priced at $10 per 1 million input tokens and $20 per 1 million output tokens. For text, input tokens cost $0.60 per 1 million, while output tokens are $2.40 per 1 million. Cached audio and text are both available at $0.30 per 1 million tokens.

Enhanced Control for Voice Interactions

Developers now have greater control over voice interactions, allowing developers to fine-tune how and when AI responses are delivered during voice interactions, with the following features:

• Concurrent out-of-band responses: Run background tasks like content moderation or classification without disrupting the user’s voice experience.

• Custom input context: Specify which parts of the conversation to include as model input. For example, focus on a user’s last message for moderation checks or reuse a previous response without permanently altering the session state.

• Controlled response timing: Use server-side Voice Activity Detection (VAD) to gather additional information—such as account details—before manually triggering a response, allowing for improved timing and accuracy.

• Extended session length: Sessions can now last up to 30 minutes, doubling the previous 15-minute limit.

Preference Fine-Tuning: Customizing Models to Your Needs

OpenAI has introduced Preference Fine-Tuning, a new method for customizing AI models based on specific user and developer preferences. Unlike traditional Supervised Fine-Tuning—which trains a model to replicate labeled outputs—Preference Fine-Tuning teaches the model to favor preferred responses over less desirable ones.

This method is particularly effective for tasks where “better” responses are subjective, such as creative writing or summarization. Developers can use Preference Fine-Tuning to:

• Adjust the tone, style, and behavior of responses.

• Improve performance on subjective tasks through A/B testing or human feedback.

  For example, Rogo AI, a company developing AI tools for financial analysts, used Preference Fine-Tuning to improve their AI assistant’s ability to break down complex queries into sub-queries. While Supervised Fine-Tuning struggled with 'out-of-distribution query expansion'—such as missing metrics like ARR (Annual Recurring Revenue) for questions like 'how fast is company X growing'—Preference Fine-Tuning resolved these challenges. By leveraging their expert-built benchmark, Rogo-Golden, Rogo AI improved performance from 75% accuracy in the base model to over 80%.

Preference Fine-Tuning begins rolling out today for gpt-4o-2024-08-06 and will soon be available for gpt-4o-mini-2024-07-18. It will be offered at the same price per trained token as Supervised Fine-Tuning, with support for newer models expected early next year.

New Go and Java SDKs: Expanding Language Support

To make its APIs more accessible, OpenAI has released official SDKs for Go and Java, joining existing support for Python, Node.js, and .NET. These SDKs provide developers with tools to easily integrate OpenAI models into their applications:

• Go: Ideal for backend systems and APIs, the Go SDK simplifies development in highly concurrent applications.

• Java: A staple in enterprise software, the Java SDK enables seamless integration with robust typed requests and responses. The OpenAI Java SDK offers typed request and response objects along with useful utilities to simplify API request management.

Developers can find the beta SDKs on GitHub and start experimenting with OpenAI models in their preferred programming language. You can find the Go here, and the Java here.

What This Means for Developers

OpenAI’s latest tools and updates make it easier, faster, and more cost-effective for developers to build AI-driven applications. The o1 model provides advanced reasoning capabilities with state-of-the-art performance, while the Realtime API unlocks real-time voice interactions with reduced costs. Tools like Preference Fine-Tuning allow for deeper model customization, ensuring AI can meet the unique needs of developers and businesses.

For developers experimenting with generative AI, robotics, or voice-based tools, OpenAI’s updates offer significant advancements. By expanding language support and improving model flexibility, OpenAI is equipping developers with the tools they need to push the boundaries of what AI can achieve.

Day 8: ChatGPT Search With Citations

OpenAI has revealed ChatGPT Search, a game-changing feature that merges conversational AI with web search, as part of its "12 Days of Shipmas" initiative. Introduced in October 2024, ChatGPT Search now has expanded to enable users to receive fast, timely answers directly from relevant web sources, transforming how people interact with online information.

What ChatGPT Search Offers

This new functionality combines the natural language processing power of ChatGPT with the precision of up-to-date web search, providing:

• Timely Answers: From sports scores, news, and weather to stock quotes, ChatGPT can fetch real-time information from the web.

• Enhanced User Control: Users can let ChatGPT automatically decide when to search or manually activate it via a dedicated web search icon.

• Citations and Sources: Responses include links to high-quality sources, allowing users to explore deeper into the content.

For example, users can ask complex, conversational questions, and ChatGPT will refine answers based on follow-ups and the broader context of the chat.

Availability

ChatGPT Search is now available for:

• Plus and Team Users: Immediate access via ChatGPT’s desktop, mobile apps, and chatgpt.com.

• Enterprise and Educational Users: Rollout in the coming weeks.

• Free Users: Expected availability over the next few months. Additionally, a Chrome extension lets users access ChatGPT Search directly from the browser's URL bar.

Collaborative Partnerships with News and Data Providers

OpenAI has partnered with major publishers and data providers to integrate high-quality, trustworthy information into ChatGPT Search. Collaborators include:

• News Partners: The Associated Press, Reuters, Vox Media, Le Monde, Prisa (El País), News Corp, Axel Springer, Time, Condé Nast, Dotdash Meredith, Financial Times, GEDI, Hearst, and The Atlantic.

• Data Enhancements: Improved visuals for weather, maps, stocks, and sports results.

Publisher Feedback:

• Pam Wasserstein, President of Vox Media: “ChatGPT search promises to better highlight and attribute information from trustworthy news sources, benefiting audiences while expanding the reach of publishers like ourselves who produce premium journalism.”

• Louis Dreyfus, CEO of Le Monde: “We are convinced that AI search will be, in a near future and for the next generations, a primary way to access information, and partnering with OpenAI positions Le Monde at the forefront of this shift. It allows us to test innovations at an early stage while safeguarding journalism’s core values and integrity.”

• Mathias Sanchez, SVP of Axel Springer SE: “As AI reshapes the media landscape, Axel Springer’s partnership with OpenAI opens up tremendous opportunities for innovative advancements. Together, we're driving new business models that ensure journalism remains both trustworthy and profitable.”

Through these partnerships, publishers gain new opportunities to reach audiences while maintaining editorial integrity.

How It Works

ChatGPT Search is powered by a fine-tuned version of GPT-4o trained with synthetic data generation techniques. It relies on third-party search engines and publisher-provided content to deliver reliable, actionable answers. Learn more here.  

Planned Enhancements Include:

• Expanding search to shopping and travel topics.

• Extending capabilities to Advanced Voice, canvas, and Free or logged-out users in the future.

Why It Matters

This innovation marks a shift in how users access and engage with online content:

• Time-Saving and Conversational: ChatGPT simplifies search by removing the need for multiple queries and link-hopping, making information gathering faster and more intuitive.

• Publisher Benefits: By integrating citations, ChatGPT Search ensures publishers retain visibility and attribution, balancing user convenience with content creators’ needs.

• Future of AI Search: As AI continues reshaping the media and search landscape, tools like ChatGPT Search highlight how conversational interfaces can democratize access to high-quality information.

Looking Ahead

ChatGPT Search represents a significant step forward in OpenAI’s mission to integrate AI into everyday tasks. As feedback rolls in, OpenAI plans to refine the tool further, expanding its functionality while maintaining partnerships with leading publishers.

With its real-time web integration and conversational interface, ChatGPT Search is poised to redefine how users access and consume information in a digital-first world.

Day 7: ChatGPT Launches Projects: Organize and Customize Your Conversations

On the seventh day of Shipmas, OpenAI unveiled Projects in ChatGPT, a highly anticipated feature that allows users to organize, customize, and streamline their ChatGPT experience. This powerful addition brings organization to the forefront, enabling users to group conversations, upload files, and set tailored instructions for specific workflows—all while integrating seamlessly with ChatGPT’s core capabilities like Search and Canvas.

Rolling out today to Plus, Pro, and Teams users, Projects is designed to enhance productivity, foster creativity, and provide a more personalized ChatGPT experience.

Key Features of Projects

Projects offer a range of tools to improve organization and customization:

• Custom Instructions: Users can set detailed instructions tailored to each project, helping ChatGPT adapt its responses to specific needs.

• File Uploads: Attach relevant files directly to projects for easy reference in conversations.

• Conversation Organization: Group and manage related chats into smart folders for streamlined workflows.

• Search Integration: Search past conversations and add relevant ones to projects for easy access and continuity.

• Canvas Support: Use Canvas for drafting emails, documents, or code within a project’s context.

Live Demos Highlight Real-World Applications

During the live announcement, the OpenAI team showcased how Projects can be used for a variety of tasks, from managing a Secret Santa gift exchange to creating a home maintenance log and even refining personal websites:

• Secret Santa Organization: Files like survey results, rules, and event details were uploaded to a festive project. ChatGPT used this data to assign gift-givers and help find personalized gift ideas online.

• Home Maintenance Logs: By uploading files like appliance manuals and maintenance records, users could ask ChatGPT questions like, “Do I need to replace my fridge water filter?” and get tailored answers based on their specific data.

• Website Revamps: Developers demonstrated how Projects could manage code templates, personal information, and project files to streamline website updates and incorporate features like testimonials.

Why Projects Matter

The introduction of Projects represents a major leap in how users interact with ChatGPT, offering:

• Improved Workflow Management: Group conversations and files by topic for better organization.

• Tailored Interactions: Use custom instructions to create project-specific chat experiences.

• Versatility Across Tasks: From personal organization to professional projects, Projects caters to diverse needs.

By blending these features with existing tools like Conversation Search and Canvas, OpenAI has created an ecosystem that promotes efficiency and creativity.

Availability and Future Plans

Projects are rolling out today for Plus, Pro, and Teams users, with availability for free users coming soon. Enterprise and EDU users can expect access early next year.

This feature reflects OpenAI’s commitment to improving the core ChatGPT experience and addressing user feedback. As highlighted during the announcement, Projects is part of a broader effort to make ChatGPT more customizable and user-friendly.

Looking Ahead

The introduction of Projects marks a pivotal moment for ChatGPT, bridging the gap between simple chat interactions and robust project management. By enabling users to customize and organize their workflows, OpenAI has set the stage for more productive and efficient uses of AI.

This innovation also highlights the evolving landscape of AI tools, where personalization and integration are key to driving user adoption. With Projects, OpenAI demonstrates its dedication to creating practical, user-centered features that empower individuals and teams alike.

Day 6: OpenAI intorduces advanced voice with video & Santa mode

The headline feature of Day 6 is the addition of video and screen sharing capabilities to Advanced Voice Mode, making ChatGPT even more interactive and collaborative.

Key Features and How It Works:

Video Chat:

• Users can now communicate with ChatGPT via a live video interface, making interactions feel more personal and engaging.

• This feature allows for dynamic learning experiences, such as following step-by-step tutorials or engaging in real-time Q&A sessions with a visual component.

For example, the demo showed ChatGPT guiding a user through making pour-over coffee while recognizing the tools on screen and providing real-time feedback.

Screen Sharing:

• Share your screen during conversations to collaborate or troubleshoot with ChatGPT.

• The screen-sharing feature allows ChatGPT to analyze visible content (e.g., emails, messages, or workflows) and provide tailored suggestions or assistance.

A live demo illustrated ChatGPT helping a user craft a polite response to a message by analyzing the content on-screen.

Enhanced User Experience:

• Conversations in Advanced Voice Mode are powered by ChatGPT’s multimodal GPT-4 Turbo model, which enables natural audio input and output with emotional tone and nuanced pacing.

• The feature supports over 50 languages, ensuring accessibility for users worldwide.

Rollout Timeline:

The feature is being rolled out starting today on the latest mobile apps and will be available to:

• Pro, Pro+, and Teams users within a week.

• Enterprise and EDU plans early next year.

• For European users subscribed to Pro plans, the rollout will take a bit longer but is actively in progress.

Special Santa Chat

ChatGPT is bringing holiday cheer with a fun and festive Santa Claus mode, available globally wherever voice mode can be used. This feature allows users to engage in real-time conversations with Santa himself, using his iconic jolly voice.

What You Can Do with Santa Mode:

• Ask Santa about his favorite Christmas traditions or which reindeer he loves most.

• Enjoy Santa's jokes and anecdotes, such as his answer to "What’s every elf’s favorite music? Rap music!"

Use it as an engaging activity for kids to learn more about Santa's life at the North Pole or listen to holiday stories.

Accessing Santa Mode:

• Find the snowflake icon on the home screen or select Santa from the ChatGPT settings page.

• Santa mode works across desktop, mobile apps, and web versions of ChatGPT with voice mode enabled.

Extra Holiday Bonus:

To ensure everyone gets a chance to enjoy Santa, ChatGPT will reset Advanced Voice usage limits once for users during their first interaction with Santa. Even after usage limits are exceeded, you can continue chatting with Santa in standard voice mode.

Looking Ahead

The rollout of video and screen sharing in Advanced Voice Mode significantly enhances how users interact with ChatGPT, making it a more versatile tool for troubleshooting, learning, and collaboration. Whether it’s helping you refine a message or teaching you a new skill like making coffee, this feature brings a new level of interactivity.

Adding Santa Mode brings a touch of seasonal magic, showing how AI can create both meaningful and playful experiences for users of all ages. With these updates, ChatGPT continues to push the boundaries of what AI-powered communication can achieve.

Day 5: OpenAI Reveal's ChatGPT in Apple Intelligence

OpenAI has announced a new integration of ChatGPT across Apple’s iOS, iPadOS, and macOS platforms, designed to make the AI assistant more accessible and user-friendly than ever. This collaboration with Apple introduces several new features, enabling ChatGPT to seamlessly integrate into everyday tasks on iPhones, iPads, and Macs.

Frictionless AI Assistance Across Apple Devices

The ChatGPT integration provides three key ways for Apple users to interact with the AI:

• Siri Integration: Siri now has the ability to delegate complex tasks to ChatGPT, offering advanced assistance when needed. For example, users can ask Siri to invoke ChatGPT to organize events or provide detailed recommendations.

• Writing Tools: Built into Apple’s system intelligence, ChatGPT enhances writing workflows by summarizing, refining, or even generating documents from scratch. Perfect for students, professionals, or anyone needing polished written content quickly.

• Camera Control and Visual Intelligence: On the iPhone 16, ChatGPT can analyze what the camera sees to provide insights or assist with tasks based on the visual input. Example: Identify objects or help rank participants in a festive Christmas sweater contest.

How to Set It Up

To enable the ChatGPT integration on Apple devices:

• Go to Settings → Apple Intelligence and Siri and enable the new ChatGPT extension.

• Log into your ChatGPT account or use the tool anonymously.

• You can customize permissions to confirm ChatGPT requests before sharing information.

These simple steps unlock a variety of features, making ChatGPT more accessible for Apple users.

A Closer Look at ChatGPT’s Capabilities

• Festive Fun: Organizing a Holiday Party With Siri’s help, ChatGPT can plan a Christmas party, generate a holiday playlist (including “All I Want for Christmas Is You” by Mariah Carey), or even create custom album art for the event.

• Camera-Driven Interactions: Users can use camera controls to submit images or live views for analysis by ChatGPT. Example: A Christmas sweater contest where ChatGPT ranks participants based on visual features like patterns and colors.

• Seamless Transition to the ChatGPT App: Any ongoing interaction initiated via Siri or camera controls can be continued in the ChatGPT app for deeper engagement, edits, or follow-ups.

Enhanced Mac Integration

The ChatGPT integration extends to macOS with features tailored for productivity:

• System-Wide Access: Invoke ChatGPT from any application using Siri or by pressing the command key twice.

• Document Analysis: Share PDFs, images, or other files with ChatGPT for summarization, in-depth analysis, or creating visual representations like pie charts.

• Custom Visualizations: Users can request ChatGPT to analyze complex documents and create visual aids, such as charts or infographics, directly from the content.

These features provide Apple users with powerful AI tools to simplify both work and play.

What This Means

The partnership between OpenAI and Apple signals a significant step toward frictionless AI integration, catering to users who want advanced assistance embedded within their favorite devices. By leveraging Siri, camera tools, and macOS workflows, ChatGPT becomes an even more versatile and intuitive companion.

As AI adoption continues to rise, this collaboration highlights the growing trend of bringing intelligent assistants into everyday tasks, reducing barriers and enhancing productivity. With Apple’s user-friendly design and OpenAI’s advanced models, ChatGPT is now more accessible and impactful than ever.

Day 4: OpenAI Expands Canvas to All Users with New Features

As part of OpenAI’s 12 Days of Shipmas, Day 4 brings significant updates to Canvas, a tool designed to enhance collaboration and creativity on ChatGPT. Previously available in beta for Plus users, Canvas is now rolling out to all users on the platform, including free plans.

With this launch, OpenAI introduces new features that expand Canvas’s capabilities for writing, coding, and customization, enabling a more interactive and seamless user experience.

What’s New in Canvas

OpenAI announced three major updates for Canvas:

• Canvas for Everyone: Canvas, previously in beta, is now integrated into ChatGPT’s main interface. This side-by-side tool lets users collaborate with ChatGPT on tasks such as writing stories, essays, or generating code, all within an editable document.

• Run Python Code Directly in Canvas: Users can now execute Python code within Canvas, seeing outputs like text or graphics in real time. This includes built-in syntax highlighting, debugging support, and code suggestions, making it a powerful tool for programmers and data scientists.

• Integration With Custom GPTs: Canvas is now available in custom GPTs, enabling personalized models to leverage Canvas for tasks such as creating documents or responding to complex prompts in a structured format.

Key Features of Canvas

Canvas transforms the way users interact with ChatGPT by providing a collaborative workspace. Here’s how it works:

1. Side-by-Side Collaboration Canvas introduces a split view, where users can chat with ChatGPT on one side and view a dynamically updated document on the other. Edits can be made directly in the document by both the user and ChatGPT, streamlining collaboration.

2. Enhanced Writing Tools Users can bold text, suggest edits, adjust document length, refine language complexity, and even add emojis with a single click. A built-in feedback system lets ChatGPT leave comments on specific parts of the document, connecting suggestions to the user’s work more clearly.

3. Programming Features Python code can now be written, debugged, and executed directly in Canvas. ChatGPT can analyze errors, offer corrections, and even fix bugs, providing immediate feedback through an integrated WebAssembly Python emulator. Developers can generate and visualize data, such as line charts or Sankey diagrams, without leaving the workspace.

4. Custom GPT Integration Canvas is now a default feature for new custom GPTs, enabling users to design personalized assistants that use Canvas for tasks like drafting letters or generating documents.

Use Cases for Canvas

Canvas is ideal for a variety of tasks, including:

• Storytelling and Creative Writing: Users can co-write stories with ChatGPT, make real-time edits, and enhance the final product with suggestions like emoji placement or language refinement.

• Essay Feedback: Canvas allows users to submit essays for in-line feedback, enabling ChatGPT to leave detailed comments and suggestions on specific sections.

• Code Debugging and Execution: Programmers can use Canvas to troubleshoot code, fix errors, and execute Python scripts with instant output.

• Custom GPT Applications: Integrating Canvas into custom GPTs allows users to build specialized tools, such as Santa’s letter-writing assistant, to streamline workflows.

How to Use Canvas

To get started with Canvas, follow these steps:

• Open Canvas: Click the "Canvas" button in the ChatGPT composer to create a new workspace.

• Write or Paste Content: Begin writing in the Canvas editor or paste existing text, code, or prompts for collaboration.

• Collaborate and Edit: Work side-by-side with ChatGPT, making edits directly in the document or providing feedback in the chat.

• Run Code: For coding tasks, paste Python code into Canvas, use the "Run" button to execute it, and visualize results like charts or diagrams.

• Use Custom GPTs: Enable Canvas for custom GPTs via the configuration screen, giving personalized assistants access to collaborative features.

Looking Ahead

The rollout of Canvas marks a new era of collaboration on ChatGPT, enabling users to seamlessly create, code, and customize. Whether you’re a writer, programmer, or casual user, Canvas offers tools to elevate productivity and creativity.

With Python execution, improved feedback systems, and integration with custom GPTs, OpenAI’s Canvas is more than just a workspace—it’s a collaborative partner for your most ambitious projects.

Day 3: OpenAI Introduces Sora: A Game-Changing Text-to-Video AI

OpenAI has unveiled Sora, its groundbreaking text-to-video AI model, now available to ChatGPT Plus and Pro users for Day 3 of Shipmas. Designed for creatives and storytellers, Sora empowers users to generate, remix, and enhance videos, pushing the boundaries of what’s possible in video storytelling.

What is Sora?

Sora is OpenAI's cutting-edge text-to-video AI model, designed to bring imagination to life by transforming text, images, and other assets into realistic, dynamic videos. This tool offers creators unprecedented capabilities to generate and edit videos seamlessly, making it a standout among text-to-video generation tools. Sora comes equipped with an array of powerful tools that cater to both beginners and advanced creators, enabling seamless and innovative video generation.

Key Features of Sora

• Video Generation: Create videos up to 1080p resolution, 20 seconds in duration, and in various aspect ratios (widescreen, square, vertical).

• Remix: Modify existing videos by changing elements within the scene. For example, you can replace objects or adjust the environment with simple text prompts. Users can control the extent of changes using strength options like subtle, mild, or strong.

• Blend: Combine two videos to create a cohesive new scene that merges the essence of both. This tool helps users craft imaginative outputs, blending visual elements from different videos into a harmonious result.

• Recut: Trim and reconfigure videos by isolating specific scenes and extending or seamlessly looping them. Recut allows users to refine their clips with precision, building continuity or entirely new sequences from existing videos.

• Storyboard: A timeline-based tool for directing videos with multiple actions or scenes. Users can define environments, characters, and actions across a sequence of frames, giving them granular control over video narratives. Storyboard also supports uploading images to kickstart video creation.

• Loop: Turn any generated video into a continuous loop by selecting start and end points. Sora fills in the gaps to create smooth transitions for endless playback. Perfect for creating hypnotic or atmospheric visuals.

• Presets: Save and apply reusable styles, aesthetics, or themes to multiple video generations. This feature is ideal for maintaining a consistent creative vision across projects or experimenting with pre-designed looks like "Stop Motion" or "Balloon World."

• Image-to-Video Transformation: Convert still images into dynamic video sequences.

Availability and Pricing

Sora is included at no additional cost for ChatGPT Plus & ChatGPT Pro users.

ChatGPT Plus Plan ($20/month):

• Allows users to generate videos of up to 5 seconds in duration.

• Resolutions are capped at 720p.

• Up to 50 priority videos (1,000 credits).

ChatGPT Pro Plan ($200/month):

• Supports videos of up to 20 seconds in duration and 5 concurrent generations.

• Resolutions go up to 1080p.

• Up to 500 priority videos (10,000 credits).

• Download without watermark.

OpenAI also offers free access to Sora’s Explore Feed, available to both ChatGPT Plus and Pro users. This feature allows subscribers to view and draw inspiration from the community’s creations, showcasing the full potential of Sora’s video generation capabilities.

Please note: Accessing Sora.com alone does not grant the ability to generate videos; users must have an active ChatGPT Plus or Pro subscription to unlock Sora's video generation features.

Sora Turbo: Faster, Smarter, and More Accessible

The newly launched Sora Turbo model improves upon its February prototype with significantly faster processing and enhanced capabilities. Users can now generate more detailed and complex videos at reduced costs, making the technology accessible to a broader audience.

Responsible Use and Safety Measures

OpenAI emphasizes safety and transparency with Sora, incorporating features like:

• Metadata with C2PA: Clearly marks Sora-generated videos for authenticity.

• Visible Watermarks: Ensures transparency in generated content.

• Deepfake Protections: Limits uploads of human images and blocks misuse, such as generating harmful or abusive content.

• OpenAI’s Sora System Card outlines safeguards and ethical considerations, and the team plans to iterate on moderation tools based on user feedback.

Pioneering Creativity with Sora - Industry Thoughts and Examples of Use

Sora is already being used by artists and animators to transform their creative visions into reality. Tools like remix, storyboard, and presets allow users to experiment with styles, blending reality and imagination seamlessly.

Sora empowers creators to break new ground in storytelling, blending imagination and reality to craft deeply personal and innovative narratives.

Nik Kleverov, Chief Creative Officer at Native Foreign and one of Sora’s Alpha Users, shared his personal experience:

“Aside from what this means for brand storytelling and agency work, I've found Sora as a great tool for me to be able to tell a very personal story about my father, who was a nonconformist artist in Soviet Russia. This is a prime example of something that probably wouldn't have been made before AI tools.”

His film came out earlier this year and screened at TIFF - Toronto International Film Festival. You can watch the video he made with Sora here.

Additionally, Nik spoke to the Los Angeles Times about this called, “OpenAI’s controversial Sora is finally launching today. Will it truly disrupt Hollywood?” You can read that article here.

Here are a few other examples of creators using Sora:

• Minne Atairu: An interdisciplinary artist leveraging Sora to craft surreal storytelling experiences. Watch Minne’s work.

• Vallée Duhamel: A visual artist blending reality and imagination to create mesmerizing video art with Sora. See Vallée’s creation.

• Lyndon Barrois: An animator using Sora to forge new paths in creativity, transforming traditional techniques into dynamic, AI-enhanced animations. Explore Lyndon’s project.

What This Means

With Sora, OpenAI solidifies its role as a leader in generative AI for visual media. By combining technical sophistication with intuitive tools, the model democratizes video creation, enabling users to tell compelling stories and explore new creative frontiers.

As video generation AI evolves, Sora’s early release allows users and developers to co-develop norms and safeguards while exploring the transformative potential of this technology.

Day 2: OpenAI Launches Reinforcement Fine-Tuning Research Program

OpenAI has announced the second day of its Shipmas celebration with the expansion of its Reinforcement Fine-Tuning Research Program. This initiative is aimed at researchers, universities, and enterprises seeking to create expert AI models fine-tuned for complex, domain-specific tasks.

What Is Reinforcement Fine-Tuning?

Reinforcement Fine-Tuning is a new model customization technique that allows developers to enhance AI performance on highly specialized tasks. By using curated datasets of dozens to thousands of high-quality tasks, developers can guide models by grading their responses against reference answers. This process reinforces the AI's reasoning capabilities, enabling it to excel at solving problems in a particular domain with greater accuracy.

Who Should Apply?

OpenAI invites research institutions, universities, and enterprises with narrow, complex tasks that require expert-led AI assistance. Ideal applicants come from fields such as:

• Law

• Insurance

• Healthcare

• Finance

• Engineering

Reinforcement Fine-Tuning works best for domains where tasks have objectively correct answers that experts can agree upon.

What Does the Program Offer?

Participants will gain alpha access to OpenAI’s Reinforcement Fine-Tuning API, providing an opportunity to:

• Test the technique on their domain-specific tasks.

• Offer feedback to refine the API before its public release.

• Collaborate with OpenAI by sharing datasets (optional) to improve model performance.

Spots for this program are limited, and OpenAI expects to make Reinforcement Fine-Tuning publicly available in early 2025.

Interested? Apply here: Reinforcement Fine-Tuning Research Program

Stay tuned as OpenAI continues rolling out announcements and advancements throughout Shipmas!

Day 1: OpenAI Unveils Full Version of o1 and ChatGPT Pro

The first day of "Shipmas" started strong, with OpenAI announcing the release of the full version of its advanced reasoning model, o1, and the introduction of ChatGPT Pro, a premium plan designed for professionals and researchers.

o1 Reasoning Model: Smarter, Faster, and Multimodal OpenAI’s o1 reasoning model, initially previewed in September, is now available in full to ChatGPT Plus and Team users. It represents a significant leap forward for users who rely on AI to solve complex problems, offering:

• Multimodal Functionality: The ability to process images and text simultaneously, making it more versatile for real-world use.

• Increased Speed and Accuracy: Evaluations show it responds to queries up to 50% faster and makes 34% fewer major mistakes compared to its preview version.

• Improved User Experience: Smarter “thinking” capabilities mean it delivers better, more detailed answers for intricate problems, whether in coding, data science, or research.

During a livestream, researchers demonstrated its multimodal abilities by solving a challenging design problem for a hypothetical space data center, producing a solution complete with equations and detailed analysis in about 10 seconds. You can visit their page to see the benchmarks & performance metrics.

For everyday tasks, o1 is also more efficient. The latest version can answer simpler questions (like listing Roman emperors) much faster than before, making it useful for both advanced and routine tasks. Altman described o1 as faster, more intelligent, and more user-friendly compared to the preview version.

Pro users can enable this feature by selecting "o1 Pro Mode" in the model picker and submitting their query. While responses may take longer to generate, ChatGPT provides a progress bar to track completion and will send an in-app notification if you navigate to another conversation in the meantime.

ChatGPT Pro: Advanced Features for Professionals ChatGPT Pro is OpenAI’s new premium tier, priced at $200 per month, and tailored to those who need research-grade intelligence daily. This plan includes:

• o1 Pro Mode: A version of the o1 model that leverages additional computing power to deliver even more accurate and reliable answers, especially for complex questions.

• Unlimited Access: Includes the latest versions of o1, o1-mini, GPT-4o, and Advanced Voice capabilities, ensuring maximum versatility.

For researchers, engineers, or small business owners, this plan provides access to cutting-edge tools designed to boost productivity and tackle advanced challenges.

ChatGPT Pro Grants: Supporting Transformative Research To further its mission of benefiting humanity, OpenAI announced ChatGPT Pro Grants for researchers working on critical issues in medicine and science. The initial grants were awarded to experts focused on:

• Rare disease discovery: Identifying genes linked to rare and orphan diseases.

• Aging and dementia research: Advancing the understanding of aging-related cognitive decline.

• Cancer immunotherapy: Innovating new treatments for cancer using genomic medicine.

Recipients include leading scientists from institutions such as Harvard Medical School, Boston University, and Berkeley Lab. OpenAI plans to expand these grants in the future to support more groundbreaking work in other fields.

Quote from Sam Altman: “It'll be a way to show you what we’ve been working on and a little holiday present from us,” OpenAI CEO Sam Altman said during Thursday’s livestream.

Looking Ahead: What to Expect from Shipmas OpenAI’s “Shipmas” campaign is setting a festive tone while showcasing its latest advancements in AI. By releasing daily updates, the company keeps enthusiasts, leaders, and small business owners engaged with a steady stream of innovations.

With more announcements to come, the excitement is building around what’s next—whether it’s groundbreaking features, new tools, or creative demos. This campaign reflects OpenAI’s commitment to advancing AI while engaging a broad audience with accessible, impactful updates.

Editor’s Note: This article was created by Alicia Shapiro, CMO of AiNews.com, with writing, image, and idea-generation support from ChatGPT, an AI assistant. However, the final perspective and editorial choices are solely Alicia Shapiro’s. Special thanks to ChatGPT for assistance with research and editorial support in crafting this article.