LLM/AI Changelog

GPT-5.4 is OpenAI's latest model combining improved reasoning, coding, and agentic capabilities, with a "Thinking" mode that shows its reasoning upfront so users can redirect it mid-response.
For brand visibility, the two most relevant changes are that better deep web research could favour brands with stronger recent citation footprints, and improved context retention may mean brand associations are held more consistently across longer conversations.

A minor polish update to ChatGPT focused on feel rather than capability - more accurate answers, better web search results, and a more natural conversational tone with less unnecessary hedging and filler. Essentially OpenAI tuning the day-to-day experience based on user feedback rather than chasing benchmark scores.

Projects now make it easy to build a living knowledge base by adding sources wherever your context already lives; from apps, conversations, or quick text inputs.
Sources from apps:Paste a link to a Slack channel or Google Drive file or folder to add it directly as a project source.Additionally, use connected apps directly in chats.Sources from chats:Save useful ChatGPT responses into your project so great outputs become reusable knowledge.Ad-hoc text sources:Paste notes, briefs, or reference material directly to capture context instantly.
Learn more aboutprojects in ChatGPT.

Declarative agents upgraded to GPT‑5.1
Microsoft 365 Copilot declarative agents now use the GPT‑5.1 model with automatic model selection for performance and reasoning.
Detailes:
What changed:
Declarative agents now run on GPT‑5.1 using “auto” architecture, which selects the best model (fast or reasoning‑optimized) based on the user’s prompt. This improves speed and quality. Note that as the agent uses the "Think deeper" mode to generate higher quality responses, users will experience a slightly higher latency.
Why:
The upgrade ensures declarative agents provide more accurate and detailed responses.
Try this:
Use a declarative agent to process complex queries.
Test scenarios that require fast responses versus deeper reasoning.
Why this matters:
Business impact:
Supports higher‑quality automation and decision‑making.
Personal impact:
Provides users with faster and more accurate responses.

Gemini 3.1 Pro: A smarter model for your most complex tasksWhat:Building on the Gemini 3 series, we're releasing 3.1 Pro, a smarter and more capable model for complex problem-solving. 3.1 Pro is designed for tasks where a simple answer isn't enough, taking advanced reasoning and making it useful for your hardest challenges. This improved intelligence can help in practical applications – whether you're looking for a clear, visual explanation of a complex topic, a way to synthesise data into a single view or a step-by-step plan bringing an ambitious creative project to life.3.1 Pro is rolling out globally to the Gemini app, with higher limits for users with Google AI Pro and Ultra plans. To access, select 'Pro' in the model drop-down.Why:We want to make Gemini the most helpful AI assistant for everyone. With the advanced reasoning of Gemini 3.1 Pro, you can now tackle your most complex projects with greater confidence.

Web
Attach up to 20 files in a single message.You can now upload up to 20 files at once (previously 10), making it easier to analyze document sets, compare multiple files, or provide broader context in a single message.Expanded file type support.We now support additional text and code file formats, improving compatibility with source files, logs, configuration files, and other structured text.Cleaner “Select All” behavior (Cmd+A).Cmd+A now selects only the conversation content, excluding surrounding interface elements for cleaner copying and sharing of chat transcripts.
Android
Quick tools in the composer.A new quick-tools menu in the message composer makes common actions easier to access.Simplified composer.The message composer has been streamlined to reduce visual complexity and make it easier to focus on your message.
iOS
Performance improvements.We’ve made multiple performance enhancements to improve responsiveness and overall speed throughout the app.
Web, iOS, and Android
Improved reliability for long “Thinking” chats.We've reduced interruption rates in longer “Thinking” chats through infrastructure improvements.

Solve modern science, research and engineering challenges with Gemini 3 Deep ThinkWhat:We’re releasing a major upgrade to Gemini 3 Deep Think, our specialized reasoning mode, built to push the frontier of intelligence and solve modern challenges across science, research, and engineering. By blending deep scientific knowledge with everyday engineering utility, Deep Think now moves beyond abstract theory to help drive practical applications.This update is available to Google AI Ultra Subscribers in the Gemini App. To access, select “Deep Think” in the prompt bar. Once you submit your task, Gemini will let you know when your response is ready – generally in a few minutes.Why:We're committed to bringing Google's latest AI innovations faster to Google AI Ultra users, including our most powerful reasoning capabilities to help you tackle even more complex problems.

We’re introducing improvements to deep research in ChatGPT to help produce more accurate, credible reports with greater control. You can now focus research on specific websites and a larger collection of connected apps as trusted sources.A redesigned sidebar entry point and fullscreen report view make it easier to start, review, and manage research in one place. Create and edit a research plan before it begins, and track progress with the ability to adjust direction mid-run.
Rolling out throughout today to Plus and Pro users, with Free and Go users following in the coming days.For more information, please refer to our help center article:Deep research in ChatGPT

We’re making an update to GPT-5.2 Instant in ChatGPT and the API that improves response style and quality.Users should notice responses that are more measured and grounded in tone, in a way that’s more contextually appropriate to the conversation. The model also tends to output clearer, more relevant answers to advice-seeking and how-to questions, more reliably placing the most important info upfront.

We’re beginning to roll out a test for ads in ChatGPT in the U.S. for logged-in adults on the Free and Go tiers. Plus, Pro, Business, Enterprise, and Education plans do not have ads. The test will be a subset of users to start. We will expand over the coming weeks as we learn from this test.ChatGPT’s answers are generated independently from advertising. Answers are optimized based on what’s most helpful to you. When you see an ad, they are always clearly labeled as sponsored and visually separated from the organic answer.Ads are shown using information that stays within ChatGPT and is not shared with advertisers. This includes what you’re discussing, how you interact with ads (such as ads you hide or engage with), and your past chats and memories, if you have ads personalization enabled. You’re in control of this experience: you can dismiss ads, share feedback, learn how and why you’re being shown an ad, delete your ad data with one tap, and manage ad personalization at any time.Advertisers never see your chats or personal details and receive only aggregated performance information, like impressions and clicks. During the test, ads do not appear in chats about sensitive topics or regulated topics, including health, mental health, or politics, and are not shown to users who tell us or we predict are under 18.

Jan 10, 2026:We lowered the Standard and Light thinking time as we observed users prefer faster responses. As part of this update, theExtendedthinking setting for GPT-5.2 was unintentionally changed to be lower which we have now fixed.February 3, 2026:We made another small reduction to Standard thinking time based on testing.February 4, 2026:We’re restoring theExtendedthinking level for GPT-5.2 Thinking to its prior setting, correcting the inadvertent reduction from January. Extended is now back to its prior level.We periodically adjust the default thinking time for our reasoning models. These changes are based on ongoing experiments to find the best balance between answer quality and response speed for users.Thethinking level toggleintroduced in September 2025 gives users more choice beyond Standard, allowing them to select the right thinking level for their question—whether they want lighter, faster responses or more extended reasoning when depth and accuracy matter more.Thinking time is not directly comparable across different models. Each model is tuned independently to what works best for users. We’ll continue to adjust these settings as models evolve, and will keep giving users clear controls when there are meaningful tradeoffs to choose from.

Responses in ChatGPT are now more visual and easier to scan.For everyday questions, responses may include at-a-glance visuals—like checking your team’s latest stats, converting units, or doing a quick calculation.Answers now also highlight important people, places, products, and ideas. You can tap any highlight to open a side panel with key facts and trusted sources, making it easy to understand the context without asking follow-up questions.Rolling out globally on iOS, Android, web.

Copilot Chat model selector now offers GPT-5.2
[Android, Windows, iOS, Mac, Web]
Choose GPT-5.2 in Copilot Chat for faster responses or deeper reasoning, improving accuracy and clarity.
Details:
What changed:
The model selector in Copilot Chat now includes GPT-5.2. Users can pick Quick Response for immediate answers or Think Deeper for more thorough reasoning. GPT-5.2 delivers better instruction following, improved math and coding performance, and clearer explanations compared to previous models.
Why:
This update provides more flexibility and precision, helping users handle both simple and complex tasks with adaptive reasoning.
Try this:
Select More in the model selector and choose Quick Response for quick answers.
Choose Think Deeper when you need detailed reasoning for complex queries.
Why this matters:
Business impact:
Enhances productivity by offering tailored response modes for different workflows.
Personal impact:
Gives individuals control over response depth, improving efficiency and clarity.
Additional resources:
Blogs:
Available today: GPT-5.2 in Microsoft 365 Copilot

We’re updating GPT-5.2 Instant’s default personality to be more conversational and better at adapting its tone contextually, making exchanges feel smoother and more natural. You can still select a different base style and tone for ChatGPT, along with tuning characteristics like warmth and emoji use, within the Personalization menu in settings.

When reference chat history is enabled, ChatGPT can now more reliably find specific details from your past chats when you ask. Any past chat used to answer your question now appears as a source so you can open and review the original context.This memory improvement is now available for Plus and Pro users globally.

GPT-5 powers Copilot Chat by default
[Android, Windows, iOS, Mac, Web]
Copilot Chat delivers faster, smarter results using GPT-5 as the default model.
Details:
What Changed:
All Copilot Chat experiences run on GPT-5 by default. Copilot automatically routes prompts to the best-performing models for each task, using fast models for basic tasks and specialized reasoning models for multi-step requests.
Why:
Users want speed, accuracy, and minimal friction. This update helps to ensure consistent, high-quality AI interactions in Copilot Chat.
Try This:
“Get me up to speed on the latest plans related to [project/initiative]. Help me think through what to do next.”
“Use the attached spreadsheet with customer feedback to create a polished executive report that helps upper management decide where to prioritize resources in our next cycle.”
Why this matters:
Business Impact:
Increases accuracy and productivity, helping teams make strong decisions and complete complex work faster.
Personal Impact:
Removes friction by letting Copilot choose the best model--fast or reasoning--for the prompt.
Additional resources:
Blogs:
Available today: GPT-5 in Microsoft 365 Copilot

Shopping research is a new interactive experience that helps you find the right products without doing all the research yourself. Describe what you need—like a quiet cordless vacuum for a small apartment, a gift for a four‑year‑old who loves art, or a way to choose between three bikes—and ChatGPT will ask clarifying questions, research deeply across the internet, review quality sources, and build on ChatGPT’s understanding of you from past conversations and memory to return a personalized buyer’s guide with options, tradeoffs, and links to buy.
ChatGPT learns what you like through an interactive interface allowing you to guide the research based on your real-time feedback.
Shopping research is available on ChatGPT Web, iOS, and Android for logged‑in Free, Go, Plus, and Pro users. To help with holiday shopping, we’re making usage nearly unlimited across plans through the season. To learn more, visit theblog.

Gemini 3, our most intelligent model yet, is available to everyone in the appWhat:Gemini 3 Pro is rolling out to everyone in the Gemini app starting today, with higher limits for users in the Google AI Plus, Pro and Ultra plans. It brings state-of-the-art reasoning to your most complex problems, and is our best vibe coding model yet. It can also reason across text, images, audio and video better than ever before. So whether you're uploading a video of you playing a sport to get advice on how to improve your form, or transcribing notes from a lecture that you missed, the Gemini app can help without missing a beat.Starting from your first use, you'll also notice that responses are more helpful, better formatted and more concise.Try Gemini 3 Pro by selecting 'Thinking' in the model drop-down.Why:We want to make Gemini the most helpful AI assistant for everyone. With the advanced reasoning of Gemini 3 Pro, you can now tackle your most ambitious projects with greater confidence and efficiency.Introducing visual layout and dynamic view in Gemini LabsWhat:We're introducing two new experimental Labs features in the Gemini app: visual layout and dynamic view. Powered by Gemini 3 and new advances from Google Research, these features make your responses more engaging and interactive.Visual layout uses multimodal capabilities to move beyond text, generating a visually immersive response with photos and interactive modules. These elements solicit your feedback and help you further customise Gemini's response across multiple turns. Try asking Gemini to 'plan a three-day trip to Rome next summer', and you'll get a visual itinerary that you can actually explore and customise based on what interests you.Dynamic view takes this a step further using agentic coding capabilities. Gemini will design and code a unique experience with a user interface that's perfect for your specific prompt. For example, ask Gemini to 'explain the Van Gogh Gallery with live context for each piece' and you'

We’re introducing Instant Checkout in ChatGPT and the Agentic Commerce Protocol that powers it, built with Stripe.
ChatGPT can help you find anything on the web and now with Instant Checkout, you can seamlessly buy directly from the merchant, right in chat. Starting today, U.S. ChatGPT Plus, Pro, and Free users can buy directly from U.S. Etsy sellers through ChatGPT, with over 1 million Shopify merchants, like Glossier, SKIMS, Spanx and Vuori, coming soon.
Instant Checkout
Instant Checkout makes it easy to go from chat to checkout in just a few taps.
Ask ChatGPT a shopping question and it finds organic, unsponsored product results from across the web.Choose an item and ChatGPT shows where to buy, ranking merchants by factors like inventory, price, quality, primary seller status, and Instant Checkout availability.If Instant Checkout is offered, tap Buy, confirm shipping and payment details, and pay in one tap with your card on file or other express pay options.If Instant Checkout isn’t available, ChatGPT provides a direct link so you can purchase from the merchant’s site.
You’re in control at all times and give explicit confirmation to purchase. Payment goes directly from you to the merchant, not through OpenAI.Learn more about Instant Checkout.
Agentic Commerce Protocol
The Agentic Commerce Protocol (ACP) gives merchants and developers a way to build agentic commerce experiences that work with their current systems and connect them with 700M+ weekly active users on ChatGPT.
If you’re a merchant and want to add Instant Checkout in ChatGPT, apply on ourmerchant page.We’ll be accepting and onboarding merchants on a rolling basis.
Developers can start exploring the Agentic Commerce Protocoldocumentationand example implementations onGitHub.

We’ve made further improvements to search in ChatGPT, making results more accurate, reliable, and useful.
Factuality:Fewer hallucinations, improving answer quality.Shopping:Search is better at detecting shopping intent, showing products when you want them and keeping results focused when you don’t.Formatting: Answers are better formatted for quick understanding without losing detail or quality.

Personalize interactions with Copilot Memory
[Android, iOS, Web]
Copilot Memory leverages insights inferred from conversations between the user and Copilot, along with data from the Microsoft Graph and custom instructions to provide personalized help for tasks. Users have full control and can view, manage, disable or clear memory at any time.
Learn more.

GPT-5 now available in Copilot Chat
Unlock advanced reasoning in Copilot Chat with GPT-5, offering dynamic model switching to address both simple and complex queries. Seamlessly obtain fast answers or dive deep into data analysis for tasks like summarizing RFPs or evaluating detailed proposals. Empower your productivity with intelligent, context-aware solutions tailored to your needs.
Learn more.

Copilot in Excel with Python | Reasoning Model Integration (Think Deeper)
[Mac, Windows, Web]
While performing advanced analysis with Copilot in Excel with Python, users can choose the "Think Deeper" mode to get a more elaborate and detailed plan, followed by automatic execution to generate Python code, results, and explanations. This improves performance on complex asks by leveraging the power of the latest AI reasoning models.
Learn more
.