ChatGPT - Conversational AI Assistant by OpenAI | WhatAI

chatgpt.com · OpenAI · Covers all plans · Updated April 25, 2026

Tags: GPT-5.5 · Images 2.0 · File Library · Ads · Agentic work · Model picker

OpenAI has shipped more changes to ChatGPT in the past fortnight than most platforms push in a quarter. Some of it is headline-grabbing, like the GPT-5.5 model release. Some of it is quieter but arguably more useful day-to-day, like a file system that finally keeps your uploads between sessions. Below is a plain-English account of what changed and what it actually means if you use ChatGPT regularly.

900M weekly active users · 50M+ paying subscribers · 98 countries on Go plan

GPT-5.5 arrives, seven weeks after GPT-5.4

On April 23, OpenAI released GPT-5.5, its newest model, to paid subscribers across Plus, Pro, Business, and Enterprise plans. GPT-5.5 Pro is available on the higher tiers. The release came roughly seven weeks after GPT-5.4, which itself arrived only weeks after GPT-5.3. The cadence is worth noting: these are no longer once-a-year launches. They are closer to software updates.

OpenAI president Greg Brockman described the model as a meaningful step forward in how AI handles computer work. The core claim is that GPT-5.5 can take on a vague, multi-part task, figure out what needs to happen, use available tools, check its own work, and continue until the job is done, without needing step-by-step instruction from the user. The company says this is especially noticeable in coding, research, document creation, and tasks that span multiple tools or data sources.

Internally at OpenAI, the company says more than 85 percent of staff now use Codex weekly. One team used the model to review over 24,000 tax forms across more than 71,000 pages, finishing two weeks ahead of schedule compared to the prior year. Another automated weekly business reporting, saving several hours per person per week. These are internal examples, so they should be read with some skepticism, but they give a reasonable sense of the kinds of tasks the model is being aimed at.

"It can look at an unclear problem and figure out just what needs to happen next." — Greg Brockman, OpenAI president

One technical detail worth knowing: GPT-5.5 is priced higher than GPT-5.4 for API access, but OpenAI says it uses fewer tokens to complete the same tasks, which may offset costs in practice. For ChatGPT subscribers, pricing is unchanged.

The model picker is simpler now

Alongside GPT-5.5, OpenAI has cleaned up how users select models inside ChatGPT. Instead of a growing list of numbered versions, the options are now Instant, Thinking, and Pro. Instant covers fast everyday tasks and defaults to GPT-5.3. Thinking covers harder problems that benefit from longer reasoning, using GPT-5.5. Pro is the highest-capability option for complex, long-running work. The system can also switch automatically: if you pick Instant but ask something complicated, ChatGPT may quietly route to a stronger model.

GPT-5.1 was retired in March 2026. If you had conversations using that model, they have been moved to the nearest current equivalent without any action needed on your part.

Images 2.0 and the thinking mode for visuals

On April 22, OpenAI introduced ChatGPT Images 2.0. The new image engine handles a wider range of aspect ratios and is notably better at rendering text within images, which has historically been a weak point for AI image generation. There are two versions: a standard model available to all users, including those on the free plan, and a thinking mode that applies additional reasoning before generating an image. The thinking mode takes longer but tends to produce cleaner results on complex prompts.

File Library: your uploads now persist between sessions

This one arrived in late March but is still rolling out and worth covering here. The File Library lets ChatGPT save files you upload or create, so you can come back to them in future sessions without re-uploading. PDFs, spreadsheets, images, notes, and drafts are all supported. Previously, anything you uploaded disappeared when a conversation ended. For people doing ongoing research, editing projects, or anything that involves working with the same documents repeatedly, this removes a real point of friction.

There are caveats. Temporary Chat sessions do not save files. Availability varies by plan, device, and region. The experience is most complete on the web version of ChatGPT, with mobile support more limited at launch.

What changed by plan

• Free: Access to Images 2.0 standard and GPT-5.4 mini via the Thinking option. Ads are now appearing in Australia, New Zealand, and Canada.
• Go: The plan expanded to 98 countries. Includes more messages, larger uploads, and GPT-5.4 mini for Thinking tasks. Ads apply in the same regions as Free.
• Plus: GPT-5.5 Thinking is available. File Library is rolling out. Images 2.0 with thinking mode is included.
• Pro: GPT-5.5 Pro is available. Early testers report notably stronger output on legal, business, data science, and scientific tasks.
• Business / Enterprise: Updated integrations for Box, Notion, Linear, and Dropbox with write capabilities. Outlook shared mailbox support. Full MCP connector publishing for custom workflows.
• Students: Codex credits available at chatgpt.com/codex/students via university email verification. Credits are valid for 12 months from activation.

The ad rollout and what it covers

OpenAI has started showing ads to Free and Go users in Australia, New Zealand, and Canada. The ads are targeted using information that stays within ChatGPT, including what you are discussing and your interaction history. That information is not shared with advertisers, who receive only aggregated data like impressions and clicks. Ads do not appear in conversations about health, mental health, or political topics, and are not shown to users identified or predicted to be under 18.

Whether this test expands to other regions is not confirmed. OpenAI has not announced a timeline for broader rollout.

Integrations: what got updated for teams

For Business and Enterprise users, several connected apps have been updated with write capabilities. Box, Notion, Linear, and Dropbox now support actions beyond reading, meaning ChatGPT can create or update content in those tools directly, where workspace admins have enabled those actions. Outlook Email and Calendar integrations now cover shared mailboxes and delegated calendars. Google Drive has been consolidated into a single app connection covering Docs, Sheets, and Slides.

MCP (Model Context Protocol) support is now available with developer mode for Enterprise workspaces, allowing teams to build and publish custom connectors that give ChatGPT the ability to take actions in internal tools.

The picture that emerges from these updates is a platform shifting from question-and-answer toward something more like a persistent work environment. The File Library, the agent-focused model improvements, the tighter integrations with productivity tools: each of these moves ChatGPT closer to a place where you do work rather than just ask things. Whether that works as well in practice as it does in release notes is something users will figure out in the weeks ahead.

Sources: OpenAI release notes (help.openai.com) · openai.com/index/introducing-gpt-5-5 · Fortune · TechCrunch · Axios · 9to5Mac · CNBC · Published April 25, 2026 by the WhatAI editorial team

ChatGPT is a conversational AI assistant developed by OpenAI. It uses large language models to process natural language inputs and generate responses. The system follows instructions, maintains context across messages, and adapts replies based on the ongoing conversation.

ChatGPT supports text, voice, and image inputs. It provides access to models in the GPT-5 series, including variants with extended reasoning and context windows up to 256k tokens in selected modes. Additional functions include memory of past conversations, content drafting and summarization, logical problem solving, language translation, and agentic task execution through integrated tools such as web browsing and code interpretation. The platform is accessible via web interface, mobile apps, and desktop applications with tiered usage options.

Use Cases

Pricing varies by plan and region — see current pricing.

ChatGPT Community Discussions

Explore community discussions. Ask and answer questions on ChatGPT to grow and learn together.

WhatAI Recommended Watch: ChatGPT Tutorial for Beginners – Learn How to Use ChatGPT

Watch This 27-Minute Tutorial for a Step-by-Step Introduction

At WhatAI we recommend helpful tutorials on popular AI tools for our community. This ChatGPT tutorial for beginners from Charlie Chang provides a clear guide showing how to sign up, access the free version, and use 15 practical prompts for tasks such as requesting facts, generating suggestions, translating text, making comparisons, creating stories or emails, brainstorming ideas, summarizing content, building lists, listing pros and cons, finding quotes or studies, giving feedback, role-playing scenarios, explaining concepts, and getting advice. The video covers the basics and encourages viewers to follow along hands-on. We point our community to this video because it offers a straightforward introduction to using ChatGPT. Watch the full tutorial to learn more.

ChatGPT Pros & Cons

What input types does ChatGPT support?

ChatGPT accepts text, voice recordings, and uploaded images for analysis and response generation.

Which models power ChatGPT?

ChatGPT uses models from the GPT-5 series, with options for standard responses and deeper reasoning modes depending on the selected setting.

Does ChatGPT remember previous conversations?

ChatGPT can reference details from past conversations within the same chat thread and across sessions when memory is enabled.

Can ChatGPT perform actions beyond text generation?

ChatGPT includes agentic features that allow it to use tools such as web browsing and code execution to complete multi-step tasks under user guidance.

Is ChatGPT available without a paid subscription?

ChatGPT offers a free tier with usage limits alongside paid subscription options that provide higher limits and access to advanced model features.

Related AI Models: LLMs, Multimodal Systems, and More Tools

Explore the Network