ChatGPT Just Got a Lot More Powerful. Here Is Everything That Changed in the Last 30 Days.

ChatGPT Just Got a Lot More Powerful. Here Is Everything That Changed in the Last 30 Days.

If you use ChatGPT and have not checked what it can do lately, there is a reasonable chance it can now do several things you did not know were possible. Over the past 30 days, OpenAI has shipped a set of updates that represent the most significant expansion of ChatGPT's capabilities since the platform launched.

Nearly 700 million people use the service every week. Here is a plain-language account of what changed, what it means, and what is still being rolled out.

The New Engine: GPT-5.5

The biggest change underneath all the others is the model itself.

OpenAI launched GPT-5.5 on April 23, 2026, describing it as a "new class of intelligence for real work." The model comes in three variants. GPT-5.5 Instant is now the default model for all ChatGPT users, replacing GPT-5.3 Instant. GPT-5.5 Thinking, available to Plus, Pro, Business and Enterprise users, is designed for harder problems that require more step-by-step reasoning. GPT-5.5 Pro, available only to Pro, Business and Enterprise subscribers, is described by OpenAI as offering the best overall performance for demanding tasks, with early testers describing it as "a step up in both the difficulty and quality of work ChatGPT can take on."

According to OpenAI, GPT-5.5 is stronger at coding, research, data analysis, document-heavy tasks and longer assignments than any previous version. The company says it is also more token-efficient than GPT-5.4, meaning it gets more done with less computational overhead, though its API price is double that of its predecessor.

The practical change that most users will notice: answers are cleaner, more direct, less prone to over-formatting, and the model asks fewer unnecessary follow-up questions.

Meanwhile, older models are being phased out. GPT-4.5 will be retired from ChatGPT on June 27, 2026. OpenAI o3 follows on August 26, 2026. GPT-5.2 Thinking was retired on June 5, 2026.

ChatGPT Now Manages Your Personal Finances

One of the most discussed new features of the past month is the Personal Finance experience, which OpenAI launched with a dedicated announcement.

Users with a ChatGPT Plus or Pro subscription can now connect their bank accounts, credit cards and financial accounts directly to ChatGPT. Once connected, the tool can analyse spending patterns, build monthly budgets, model mortgage repayment scenarios, review tax documents, and generate personalised financial plans.

OpenAI reported that GPT-5.5 Thinking, the default model in the Finances feature, scored 79 out of 100 on its internal personal finance benchmark across challenging finance tasks. GPT-5.5 Pro scored 82.5 out of 100.

In one example from OpenAI's own documentation, a finance team used Codex to review 24,771 K-1 tax forms totalling 71,637 pages, completing a workflow that accelerated the task by two weeks compared to the prior year.

The feature represents a significant move by OpenAI into territory previously occupied by dedicated financial planning tools and, to a lesser extent, human financial advisers handling routine household budgeting.

ChatGPT Now Helps You Find a Job

Also launched in the past month is a live job search feature inside ChatGPT.

Users can now search for job listings directly within ChatGPT. The tool surfaces live openings from sources including Indeed, Upwork and Appcast, with results described as personalised using the user's work experience and skills. Users can also upload an existing CV or create one from scratch inside ChatGPT and then download a formatted version tailored to a specific job listing.

The feature positions ChatGPT as an end-to-end job search assistant rather than simply a tool for drafting cover letters, which was its most common employment-related use case previously.

Codex Can Now Control Your Windows Computer

Perhaps the most technically significant update of the past two weeks is Codex's expansion to Computer Use on Windows.

Codex, OpenAI's AI coding agent, can now see, click and type inside Windows applications directly. Users can assign Codex a task on their Windows machine and then continue the conversation or check on progress from a phone or Mac. The Windows machine functions as the host for project files, the local browser, an app server and other tools, while the user can direct the work remotely.

The feature is positioned primarily at developers and is currently unavailable in the European Economic Area, the United Kingdom and Switzerland at launch, citing regulatory requirements.

Additional Codex updates include Goal Mode, now generally available across the Codex app, IDE extension and command line, which allows users to define an outcome and success criteria and let Codex work toward them autonomously. A new Appshots feature on macOS lets users attach a specific app window to a Codex thread so the AI understands the visual context without requiring a detailed setup prompt.

An OpenAI Model Solved a Mathematical Conjecture

On May 20, 2026, OpenAI announced that one of its models had disproved a central conjecture in discrete geometry, a branch of mathematics dealing with combinatorial properties of geometric objects.

The announcement noted the model arrived at a counterexample to a conjecture that had been open in the academic mathematics community, meaning it produced a mathematical object or proof that demonstrated the conjecture was not universally true.

OpenAI did not identify the specific conjecture in its news headline, but the claim is significant. Disproving established mathematical conjectures is a task that typically requires years of work by specialist researchers. The announcement positions AI not merely as a tool that assists mathematicians but as one capable of original mathematical discovery.

The research community's independent verification of this result is ongoing.

What Is Coming Next: GPT-5.6

Developers testing advanced environments have reported routing identifiers pointing to a model labelled GPT-5.6 in Codex infrastructure logs, according to reporting from AI News Today.

Some ChatGPT Pro users claim to have successfully routed requests to GPT-5.6 through OpenAI's Codex environments and tested it against long-context tasks. If the pattern of releases holds, GPT-5.5 to GPT-5.6 would represent approximately a 40-day cycle between major model updates, which would place a formal announcement in early June 2026.

OpenAI has not confirmed the existence or timeline of GPT-5.6. The company does not comment on unreleased models.

The Scale of Where ChatGPT Is Now

To understand the significance of these updates, the adoption numbers matter.

OpenAI reported that nearly 700 million people use ChatGPT on a weekly basis as of the GPT-5 era. OpenAI's own Q1 2026 usage data, published in late May, found that the fastest-growing workplace tasks on the platform included content creation, health-related documentation and information retrieval. The data also found that ChatGPT was being used by a broader demographic mix and in more countries than in any previous quarter.

In a separate development, OpenAI frontier models and Codex became available on Amazon Web Services on June 1, 2026, expanding enterprise access significantly.

The company was also named a Leader in enterprise coding agents by Gartner in May 2026, a category that did not exist as a formal Gartner evaluation area until this year.

What Has Not Changed

Pricing. ChatGPT Free remains available with GPT-5.5 Instant as the default model. ChatGPT Plus remains at $20 per month. ChatGPT Pro, which includes GPT-5.5 Pro access, is $200 per month. The Personal Finance feature and job search are available on paid plans.

The concerns. Researchers and regulators continue to scrutinise OpenAI's pace of deployment relative to safety evaluations. The company's Frontier Governance Framework, published May 28, 2026, is its formal response to those concerns, laying out internal commitments on how frontier models are evaluated and deployed. Independent analysts have noted the framework is voluntary and not subject to external enforcement.

The competition. OpenAI is not the only company releasing models at pace. Google's Gemini 3.5 Pro has faced early evaluations suggesting it lags behind both GPT-5.5 and Anthropic's Claude Opus 4.8 on key performance benchmarks. Chinese AI laboratories are emerging as cost-efficient competitors with a focus on open source releases that could affect the global landscape over the next 12 to 18 months.