All The ChatGPT Updates & Timeline

Other Key Features:

• Advanced STEM Capabilities: OpenAI o3-mini delivered exceptional performance in math, coding, and science, making it a strong choice for technical reasoning.
• Production-Ready Developer Features: o3-mini was the first small reasoning model to support function calling, Structured Outputs, and developer messages, providing greater flexibility for developers.
• Reasoning Effort Levels: Developers could choose between low, medium, and high reasoning effort to balance speed and accuracy according to their needs.
• Increased Rate Limits: Plus and Team users had access to 150 messages per day, tripling the previous limit of 50 messages with o1-mini.
• Integrated Search Functionality: o3-mini can now retrieve up-to-date answers with web links, marking an early step in integrating search across our reasoning models.

For the first time, a reasoning model was accessible to free users by selecting 'Reason' in the message composer or regenerating a response.

Other Key Features:

• Dedicated Browser for Task Execution - When activated, Operator opens a separate, secure browser window within ChatGPT. Users can see the actions Operator is performing in real time and can take back control at any point.
• Advanced AI Model (CUA - Computer-Using Agent) - Operator is powered by a CUA model, which combines: GPT-4o’s vision capabilities (to "see" and interpret webpages) & Advanced reasoning abilities to decide the best actions to take.
• User Confirmation for Critical Actions - Operator asks for user confirmation before completing sensitive tasks such as submitting orders, sending emails and booking appointments. 
• Security and Privacy Measures: Operator does not collect or screenshot user data. Users must manually enter sensitive details like credit card information to prevent misuse. If suspicious activity is detected, the AI pauses execution for review.

Limitations & Known Issues:

• It may struggle with complex tasks like creating slideshows or handling intricate calendar systems.
• It cannot send emails or delete calendar events at this stage.
• It might get stuck on sites with CAPTCHAs, login pages, or non-standard interfaces.
• It has usage limits, including daily and task-dependent restrictions.

Operator is OpenAI’s first major attempt at making AI-powered agents mainstream and it is still in research preview.

How to Use Tasks?

• Choose “4o with scheduled tasks” from the model picker.

• Specify what you want from ChatGPT and the timing (e.g., “Remind me to call my family every Sunday  at 5 PM”).

• Manage tasks through a new Tasks section (available on web only).

Examples:

• Daily Weather Report at a specific time.

• Reminders (renew passport, meet, birthdays).

• Entertaining (daily jokes for kids).

• Notifications on Web, Desktop, and Mobile – Users are notified when tasks are done.

• Chat or Profile Menu Task Management – Edit or cancel tasks anytime.

• 10 Active Tasks  - Users may have at most 10 scheduled tasks running.

• Smart Suggestions (User Approval Needed) – ChatGPT can suggest tasks based on conversations.

Will Free Users Get Tasks?

OpenAI did not clarify yet if Tasks will be open to free users, and thus it is likely to remain a premium feature so as to justify subscription costs ($20/month for Plus, $200/month for Pro)

Other Key Features:

• Adjustable Reasoning Time – Users could tweak compute settings (low, medium, high) for better performance.
• AGI Claims – On the ARC-AGI benchmark, o3 scored 87.5% on high compute, tripling o1’s lowest performance.
• Advanced Problem-Solving – o3 outperformed previous models in coding, math, and science, achieving near-perfect scores on multiple benchmarks.

But There Were Concerns:

• AI Safety Risks – o1 showed a tendency for deception, and o3 may amplify this. OpenAI would use “deliberative alignment” to mitigate risks.
• Compute Costs – The highest-performing setting was exceedingly expensive, making scalability a question.

A smaller, distilled version o3-mini was also teased.

Other Key Features:

• Enhanced Reasoning & Self-Verification - Unlike standard AI models, this model could fact-check itself. This reduces hallucinations and incorrect outputs but takes longer to generate responses.
• Limited Rollout to High-Spending Developers - Initially available only to tier 5 developers, requiring at least $1,000 in OpenAI spending and an account older than 30 days since the first payment.
• Post-Trained Version for Improved Performance - The API used an upgraded “o1-2024-12-17” version with refinements based on user feedback. This update enhanced response accuracy & depth, handling of programming and business-related queries, reduction in unnecessary refusals of requests

Additionally, OpenAI has introduced a new ChatGPT Pro plan at $200/month, offering unlimited access to advanced models like GPT-4o and a special o1 version for power users.

Other Key Features:

• Flexible Aspect Ratios – It supported widescreen, vertical, and square formats across platforms.
• Riff and Remix Assets - Users could edit existing assets, or merge multiple assets together to create interesting content.
• New Version - Sora Turbo was optimized for efficiency and quality compared to the original research version released in Feb 2023.
• Global Rollout (Outside EU, UK, and Switzerland) – Launched to all other regions where ChatGPT was supported.

Future Plans
Pricing Model in 2025 - Early in the next year, OpenAI aimed to provide tiered pricing for various user types.
Competing with Industry Giants - Sora would compete against heavyweights like Google, Meta, and Stability AI to create the best AI video tool.

How to Access Sora?
Free for ChatGPT Plus and Pro users.
Added directly into ChatGPT, making video creation slow and simple without the use of any third-party tools

Other Key Features:

• Multimodal Reasoning - Processed both text and image prompts with higher accuracy
• Cost Effective – Optimized for cost-effective high-quality responses and great for developers and businesses.
• Advanced Function Calling – Enriched interaction with external systems, facilitating AI-powered automation.
• Multiple Languages – Offered the same extensive range of languages as GPT-4o.
• Improved Long-Context Management - Handled long conversations and retained context across messages significantly better than GPT-3.5 Turbo

GPT-4o Mini became the default model for any ChatGPT user on the free tier instead of GPT-3. 5 Turbo. It was also available for developers through the API that provided an inexpensive way to add AI to applications.

Other Key Features:

• Fast response times – Latency significantly shortened for relatively seamless interactions.
• Improved text comprehension & generation – Responses that were more contextually aware, coherent, and nuanced.
• Image Processing – Better at processing images; could understand, analyze, and generate images.
• Voice conversation (coming soon) - Future upgrades would allow you to talk back and forth with AI right away, making AI interaction feel more like a natural and logical method for communicating.
• Availability & Usage Limits - Free Users got limited messages on GPT-4o. Plus Users ($20/month) received 5x the message limit of free users. Team & Enterprise Users got even higher limits for business and professional use cases.

Text and image abilities were live, while role-based voice features that will let users converse with ChatGPT were expected to be rolled out soon.

Other Key Features:

• Unlimited access to OpenAI’s models - including o1, OpenAI’s most powerful “reasoning” model, tailored for complex tasks.
• Voice Features Improvements – This included the enhanced capabilities associated with the voice features that made ChatGPT more intuitive and responsive to voice interactions.
• Higher Usage Limits – As opposed to the free and Plus tiers, Pro users could max out on ChatGPT without any limitations.
• Pricing - ChatGPT Pro was $200/month, which was much more expensive compared to the Plus plan at $20/month.

While the cost may be prohibitive compared to a human being for most people, the o1 model's unique access and deduction power was worth it for businesses, researchers, and heavy AI users.

Other Key Features (GPT-4 Turbo):

• Broader Context Window – GPT-4 Turbo could read a more significant amount of data, making it highly suitable for long conversations and documents.
• Cheaper & Quicker – OpenAI had further improved the model to decrease use spending and augment the speed of answers.
• Fresh Knowledge – Although GPT-4’s knowledge had a cutoff, GPT-4 Turbo learned continually as it got better data and more capability. 

Other Key Features (Custom GPTs):

• No Code - Users could create AI assistants for specific tasks with an easy, guided interface.
• Specialized Knowledge & Instructions – You could adjust the AI’s responses to fit your needs and provide specialized data (for example, with Industry Data).
• Different Use Cases – From chatbots for customer support to AI tutors, Custom GPTs opened up new possibilities for individuals and businesses.

These additions increased ChatGPT’s pay-as-you-go, plug-and-play appeal for casual users and professionals.

Other Key Features:

• Seamless Image Creation - Type in a description for an image, be it a simple sentence or a detailed paragraph, and ChatGPT will create the image from it.
• Web and Mobile Availability – The feature was accessible from ChatGPT on desktop and mobile.
• Easy Selection – Users will now find DALL·E 3 (the image generator) as option for the GPT-4 models option.
• Capacity-Based Message Limits - The total number of image requests users could make was contingent on OpenAI’s system capacity at that time.

Business owners could use DALL·E 3 for marketing visuals, concept designs, and social media content command without having to use separate graphic design tools.

Other Key Features:

• Initial Countries – The app was first available in the US, India, Bangladesh, and Brazil.
• Global Footprint - Then, just two days later, on July 27 Open AI stated that ChatGPT for Android had expanded to countries including Argentina, Canada, France, Germany, Indonesia, Ireland, Japan, Mexico, Nigeria, the Philippines, the UK, and South Korea.
• Announced Upcoming Rollouts – Per OpenAI, more countries were going to be added.

The launch followed a trend from June in which traffic on the ChatGPT web page and app installs dropped, according to reports by Similarweb. The expansion to Android was supposed to change that trend by boosting adoption.

Other Key Features:

• Syncs Chat History – Users could continue good conversations back and forth, as the app syncs chat history to the web version.
• Validated Voice Input Using Whisper – The application utilized Whisper, OpenAI’s sophisticated speech recognition model that transformed voice to text for improved ease of interaction.
• Compatible with iPhones, iPads - The app was built natively on Apple platform and offered a fluid experience on iPads and iPhones.
• No-Ad Experience – The ChatGPT for iOS app launched for free with absolutely no ads at all, unlike many free apps. 

The launch solved one of the biggest pain points for ChatGPT users - easy access from their mobile devices.

Other Key Features:

• Enhanced Creativity and Cohesion - The model performed better at producing more nuanced, creative, and cohesive text for tasks like telling narratives, writing music, or coding.
• Larger Context Processing - Able to absorb up to 25,000 words, GPT-4 improves its ability to comprehend long-form content, complex instructions, and extended dialogues.
• Improved Accuracy and Safety - OpenAI fine-tuned the model to be 40% more likely to generate factual responses and much less likely to generate disallowed or harmful output

GPT-4 was initially released to OpenAI’s paying users through ChatGPT Plus, which costs $20 a month.

Other Key Features:

• Access to ChatGPT, even in peak times - Users could use ChatGPT at any time possible without getting the ChatGPT use limit notification
• Faster response times - Subscribers got faster responses for every prompt inputed
• Early access to latest features and enhancements - ChatGPT Plus sibscribers were promised early access to all news features and improvements to ChatGPT.

This was a defining moment in OpenAI’s trajectory. It transitioned its model from a research organization into a commercially viable business but without relinquishing its original mission of making AI broadly accessible.

Other Key Features:

• Azure's AI expansion – Microsoft also expanded Azure’s AI capabilities to enable developers to build and deploy AI applications worldwide.
• New AI-Powered Experiences - OpenAI’s models was embedded into Microsoft’s consumer and enterprise products, creating new categories of digital experiences. With the Azure OpenAI Service, developers got direct access to OpenAI models they could use to build next generation AI-powered applications on Microsoft’s enterprise-grade infrastructure and AI tools.
• Exclusive Cloud Provider - In conjunction with this agreement, Microsoft Azure became OpenAI’s exclusive cloud provider, powering all OpenAI workloads (across research, products, and API services).

As a result of this multiyear investment, Microsoft and OpenAI kept pushing the envelope and redefining the potential use cases and accessibility of AI to the world at large

Other Key Features:

• General Performance - This was a more efficient and responsive version of ChatGPT’s model. The most significant change was that ChatGPT was less likely to refuse to answer certain questions unnecessarily.
• Conversation History (Gradual rollout) – A big upgrade in usability was the addition of conversation history. This feature enabled users to:
  • Save previous ChatGPT conversations.
  • Rename saved conversations for better organization.
  • Clear out conversations they didn’t need anymore.
• Daily Message Limit Experiment – In an effort to provide and manage a consistent experience and server load across different user applications, OpenAI started measuring and restricting a selected user application to a daily message limit. 

This performance update clearly showed OpenAI’s commitment to improving ChatGPT’s performance.

Other Key Features:

• Conversational Abilities – ChatGPT could maintain context over multiple interactions, making conversations feel more natural.
• General Knowledge – Since it was trained on a very large dataset, it was able to provide several informative responses regarding a variety of subjects.
• Code Generation – Developers soon grasped its potential in the trace of programming tasks through the generation of code snippets and debuggers.
• Creative Writing - Users used ChatGPT for brainstorming, writing emails and even poetry and stories.
• Free Research Preview - ChatGPT was a free public beta, where millions of people could talk to it and OpenAI used feedback to table toward future versions.

ChatGPT hit over 1 million users in just five days of its launch - a record adoption speed for a tech product. Users flooded social media with demonstrations of ChatGPT’s impressive abilities, which ranged from answering difficult questions to producing essays and jokes.