The latest ChatGPT updates include GPT-6 with interactive visual answers, audio-file uploads, mobile setup for Dots, and new ways to organise shared work with Pages and Space. Shopping tools and document scanning have also expanded. Access depends on the feature, your plan, the app you use, and your workspace settings.
If you are searching for the ChatGPT latest update, start with the task you want to improve. A new feature is useful when it helps you understand something, prepare better work or remove a repetitive step.
I have put together this guide to explain what changed, how to try the relevant features and what to check when they do not appear. It focuses on the ChatGPT product experience, including related Work and Codex features, rather than OpenAI’s wider company news.
What is the latest ChatGPT update?
The newest dated entry I reviewed in the ChatGPT release notes is the 9 October 2026 composer-predictions beta for Codex. For everyday Chat users, the main recent change is the 7 October announcement of GPT-6 with Intelligent UI.
OpenAI’s separate 5–9 October product roundup also highlights mobile Dots setup and GPT-6.1 Sol Ultrafast. That is why “the latest update” does not always refer to one feature or one app version.
| Announcement | What changed | Who should look at it |
|---|---|---|
| 9 October | Codex composer predictions | Eligible Pro users working in supported desktop threads |
| Week of 5–9 October | Mobile setup for Dots | Eligible users delegating ongoing work |
| 7 October | GPT-6 with Intelligent UI | People who want visual or interactive explanations |
| 6 October | Audio uploads | Paid users working with recordings |
| 1 October | Shopping additions and iOS scanning | People comparing products or importing paper documents |
| 29 September | Pages, Space and GPT-6.1 Sol | Teams organising work and users of Work or Codex |
Dates indicate announcements or release-note entries, not a guarantee that every eligible account received access that day.
GPT-6 makes answers more visual and interactive
OpenAI’s GPT-6 announcement describes Intelligent UI: ChatGPT can combine text, diagrams, charts, buttons and interactive tools to suit the question. It can also give a simple text answer when that is more useful.
The announced rollout began on 7 October for Plus, Pro, Business and Enterprise in Chat, with expansion to Free and Go from 8 October. OpenAI identifies Sol for those paid tiers and Luna for Free and Go. Enterprise access is subject to administrator settings.
This is a Chat update. It does not automatically change the models used in Work and Codex.
A practical prompt to try
“Compare three fictional website packages costing $500, $1,000 and $2,000. Show an interactive comparison based on page count, support and customisation. Label all assumptions and explain which customer each package suits.”
This example uses invented package details. The goal is to see whether the interface makes the differences easier to understand. Check that the labels, calculations and recommendation agree with the information you supplied.
OpenAI also says responses can begin while the model continues reasoning. Read the completed answer before relying on its conclusion, especially when further research could change the recommendation.
Audio uploads help turn recordings into usable notes
The audio-upload feature lets you attach a recording, request a transcript and ask questions about its contents. According to the official file and audio guide, audio uploads are available on paid subscriptions and workspaces, including Enterprise, but not Free at the time checked.
Supported examples include MP3, WAV and M4A, and files can be up to 512 MB. WebM and MP4 files identified as video are not supported. Availability can depend on your selected model, region, client and workspace settings.
How I would test it
Use a short recording you are authorised to upload. Then ask:
“Summarise this recording under Decisions, Unanswered questions and Proposed actions. Include an owner or deadline only when explicitly stated. Mark unclear names or statements for review.”
Compare the result with the original audio before sharing it. OpenAI notes that transcription and speaker identification can be inaccurate, with performance varying across languages.
For a small business, the useful measure is the time needed to produce a checked summary. A polished paragraph that assigns a task to the wrong person creates more work later.
Dots can now be set up from the mobile app
Dots are agents designed to carry ongoing work between conversations. The October weekly update adds mobile setup, and the current Dots guide explains that setup is supported in the iOS or Android app, desktop app and desktop web. Mobile web is not supported.
Access is gradual. Pro availability excludes the EEA, Switzerland and UK and is limited to adults. Business Premium is supported across ChatGPT’s supported regions; Enterprise, including Edu and Healthcare, requires an administrator-enabled beta.
My advice is to start with one narrow responsibility. For example, ask for a draft content brief using approved source material and specify what the finished brief must contain. Review a few results before giving the agent a broader responsibility.
The guide provides controls for rules, connected apps and pausing a dot. Those settings deserve attention because ongoing work involves more than generating a single answer.
For a fuller explanation of task design and integrations, read my guide on how to build an AI agent for your business.
Pages and Space make shared work easier to organise
Space groups pages, files and related work. A Page is an editable document that collaborators can work on with help from their own ChatGPT.
The current guide lists creation and editing for Pro, Business and Enterprise on web and desktop. Mobile supports finding, reading and sharing pages, but not editing. Managed-workspace sharing controls apply, and the documentation flags delayed availability for Business and Enterprise workspaces with Canadian or UAE data residency.
Space replaces Library for accounts with access. Projects remain separate.
A useful first Page
Ask ChatGPT to create a project brief from a conversation, then review its goals, decisions, open questions, and next steps. Assign a person to maintain the brief so it does not become another outdated document.
Sharing a Page does not expose your private chats or memory, but information written onto that Page becomes visible to its viewers. Check the content and sharing permissions together.
Shopping updates add virtual try-on and saved products
The shopping guide describes a Try on option for clothing and accessory listings. Select it, take or upload a selfie, and ChatGPT Images can generate a preview. You can also save products to Favorites or organise them in folders.
Reference photos are retained for reuse. Manage them under Settings → Personalization → Reference photos.
Treat the generated image as an illustration. OpenAI says it does not guarantee fit, size or an exact representation. Check the merchant’s measurements, product information and return policy before buying.
For ecommerce owners, my takeaway is to keep product information specific and consistent. Clear materials, measurements, variants and images help customers evaluate a product wherever they discover it. That is sensible merchandising advice, not a promise that ChatGPT will recommend a particular store.
The iOS camera can combine scanned pages into a PDF
The 1 October release notes describe a scanning option rolling out on iOS. Open the ChatGPT camera, select the three-dot menu and choose Scan. Capture multiple pages to combine them into one PDF for upload.
This can be useful for printed notes or a short handout. Before asking for a summary, check that every page is present, readable and in the correct order. A missed page can change the meaning of the document even if the final answer sounds confident.
Work and Codex updates have their own access rules
Some ChatGPT news and updates concern a specialised work experience rather than ordinary Chat. Two recent examples are worth separating.
GPT-6.1 Sol and Ultrafast
The GPT-6.1 Sol launch page describes a model for coding, computer use and professional work through Work, Codex and the API. It says the model is not yet available in Chat.
The October weekly roundup adds GPT-6.1 Sol Ultrafast in Work and Codex for Pro $500 and eligible Enterprise and Edu plans. Enterprise access is off by default and requires enablement.
If you use these tools for business, test a representative task before changing your default. Faster output is valuable only when the result still meets your requirements.
Composer predictions
Composer predictions suggest a possible follow-up message in supported Codex desktop threads. The beta is for adults with personal Pro accounts using GPT-6 Astra or GPT-6.1 Sol in local or SSH threads.
Press Tab to accept the suggestion into the composer, review or edit it, then send it. Acceptance does not send it automatically.
Generating predictions is excluded from usage limits and credit consumption during the beta. Messages you send still follow normal usage rules. I would treat each suggestion as editable text and check that it keeps the task moving in the direction you intended.
Smaller updates you may have missed
The release notes list several other changes from late September and early October. I would not build a workflow around any of them without opening the relevant help page, but they show where ChatGPT is heading.
On 2 October, Finances expanded to Free and Go users in the US, letting them connect financial accounts on web, iOS and Android. On 29 September, OpenAI added sign-in with ChatGPT for participating products, automations that start from connected-app events, and a limited preview that lets eligible Plus and Pro users let approved apps draw on their ChatGPT allowance. Security history, a log of sign-ins and changes to MFA and passkeys, arrived on 25 September under Settings → Security and login.
If you manage an account for a team, the security history entry is the one I would check first. It costs nothing to review and tells you quickly whether anything unexpected has changed.
Custom GPT retirement is an update to prepare for
The custom GPT migration FAQ schedules retirement for 11 December 2026. Affected Enterprise workspaces with an approved deferral have an 11 February 2027 date. Follow your own account or workspace notice if a different applicable timeline is communicated.
The planned replacement route is migration to plugins. Existing GPTs remain usable until retirement subject to access, and migration availability can vary.
A few more details from the FAQ are worth knowing. Existing conversations with a custom GPT stay accessible after retirement. During migration the instructions become a skill and knowledge files are copied as reference files, but the selected model does not carry over, and the original GPT becomes read-only. For Enterprise workspaces, the FAQ lists a planned end to new GPT creation on 26 October 2026, marked as subject to change.
My preparation checklist is:
- List the GPTs your work depends on and identify their owners.
- Save representative test prompts and expected results.
- Review the published instructions and reference files before migration.
- Identify integrations that need replacement.
- Test access and output with the people who will use the new workflow.
Custom actions do not transfer automatically. The FAQ also says migration uses the latest published version, not unpublished drafts, and that the replacement starts private. Do not assume migrating a GPT also installs it for everyone or preserves its sharing arrangement.
Why a new ChatGPT feature may not appear
A release announcement describes availability conditions. It does not confirm the state of your particular account.
| What you notice | What to check |
|---|---|
| Another user has the feature, but you do not | Plan, rollout stage, age or regional eligibility |
| It works on the web but not in your app | App version and supported client |
| It appears in your personal account only | Workspace or role permissions |
| A model or tool becomes unavailable | The applicable allowance and reset notice |
| A shared page opens but cannot be edited | Edit permission and whether you are using mobile |
| An integration cannot retrieve information | Connected account and source-service permissions |
For Intelligent UI specifically, the model guidance says older ChatGPT desktop apps do not support the capability and directs users to the web. It also distinguishes Instant through Extra High from Pro effort, which uses Astra and does not support Intelligent UI.
That same guidance contains older model references alongside the GPT-6 rollout information. I would use the dated announcement to understand the launch and the controls visible in my account to confirm actual access, rather than assume every paragraph describes a completed rollout.
The same page also shows the effort levels by plan, which explains many “why can’t I see it” questions. Plus includes Medium and High but not Extra High or Pro effort. Pro, Business and Enterprise include all of them, while Free and Go do not get Medium, High, Extra High or Pro and can use Think for harder questions. Intelligent UI has no separate usage quota, and on the web you can reduce the visuals by turning off “Layout and visuals” in Settings.
How to update ChatGPT
Install the latest official mobile-app version through your device’s app store. In the supported desktop app, look for an update option in the app menu; the exact wording can differ by version and operating system, and I have not verified a single path that applies to every app. On the web, there is no separate mobile-style application package to download.
An app update can provide the required interface, but it cannot unlock a feature excluded from your plan or workspace. If access still differs from the documentation, record your account type, platform and app version before contacting support.
Which new feature should you try first
I would choose based on the recurring problem you want to solve.
- If you struggle to explain options, test an interactive comparison.
- If meeting notes take too long, test audio uploads with a short authorised recording.
- If decisions are scattered across conversations, try a shared Page where eligible.
- If a task continues over several days, evaluate a narrowly scoped dot.
- If you depend on custom GPTs, prioritise migration preparation.
For more workflow ideas, explore my AI automation ideas for small businesses. Choose one task, establish the current effort and compare it with the time needed to obtain a checked AI-assisted result.
These prompts and testing methods are editorial suggestions. They are not reported benchmarks or claims of personal testing of every feature.
My advice for using the latest ChatGPT features
Pick one feature that addresses work you already do. Test it with clear inputs, review the result and record what still needs correction. That will tell you more than the length of a feature list.
For business owners, the best update is the one that makes a real process easier to complete and easier to check.
Frequently asked questions
Some changes include Free accounts, while others require paid access. The GPT-6 announcement includes a Free and Go rollout; audio uploads, Space editing and the Codex predictions beta have different eligibility rules. Check the feature you want rather than assuming all changes share one subscription requirement.
Not necessarily. The app version, model availability and account rollout are separate. Installing an update can make a supported interface available, but your plan and workspace still determine which options you can use.
A dot is designed for ongoing responsibility and can continue work between conversations. A regular conversation is a different interaction pattern. Before delegating continuing work, define the expected output and review the dot’s access and action rules.
No. The current FAQ describes a future retirement and migration process. Check the date and instructions applying to the account or workspace that owns the GPT, and allow time to test its replacement.
On the web, OpenAI’s model guidance says turning off “Layout and visuals” in Settings reduces visual elements, though some may still appear. Intelligent UI does not work with Pro effort, and older macOS and Windows desktop apps do not support it, so use the web version there.
Start with the official release notes and ChatGPT Learn’s weekly updates, then read the feature-specific help page. Keep the announcement date, access conditions and your own account experience separate when comparing reports.