Three years on from the launch that started the whole AI boom, ChatGPT is no longer just a chatbot. In 2026 it browses the web, writes and runs code, builds documents and spreadsheets, analyses your data, generates images, talks to you in real time, remembers what matters, and — increasingly — goes off and completes multi-step tasks on its own. It is the broadest, most ambitious consumer AI product on the market. It is also the most confusing to price and, on its free plan, now comes with ads. This review breaks down exactly what you get in 2026.

The models in 2026

OpenAI has tidied up its lineup since the chaos of 2025. ChatGPT now runs on the GPT-5 family, with an auto-switching router that quietly picks a faster “Instant” model or a slower “Thinking” model depending on how hard your question is. You rarely choose a model by hand any more.

Free users get GPT-5.3 Instant — fast, capable, and a huge step up from the old GPT-3.5 and GPT-4o that powered earlier free tiers. Paid users get GPT-5.5, which became the default flagship in April 2026 and folds OpenAI’s reasoning and coding advances into a single model that’s strong at long, messy, multi-part work. The very top tiers add GPT-5.5 Pro, a research-grade variant for high-stakes analysis, with a context window stretching to around a million tokens. The older GPT-4o, GPT-4.1 and o-series models were retired from the app in early 2026.

Features: the deepest toolkit in AI

This is where ChatGPT pulls ahead. No competitor matches the sheer surface area:

Agent Mode

ChatGPT can now take a goal, plan the steps, use tools, navigate websites, fill in forms and keep going until a task is done — checking its own work along the way. It is still early and needs supervision, but it points clearly at where the product is heading.

Deep Research

Give it a question and it spends several minutes reading dozens of sources, then returns a structured, cited report. It’s one of the best research tools in any AI product, though it’s rationed (10 runs a month on Plus, far more on Pro).

Codex & coding

OpenAI’s Codex coding agent — available in the browser, on desktop and now free on mobile across every plan — is among the strongest AI coding tools available, handling real repositories, debugging and multi-file changes rather than just snippets.

Everyday power features

Beyond those headliners you also get Canvas (a side-by-side writing and coding workspace), Advanced Data Analysis (a real Python sandbox for spreadsheets and charts), Vision (read and reason over images), Advanced Voice with live video and screen-sharing, the upgraded ChatGPT Images generator, Projects with their own persistent memory, cross-chat memory, scheduled tasks, and a huge library of user-built Custom GPTs in the GPT Store. OpenAI’s Atlas browser even puts an agentic ChatGPT at the navigation layer of the web. One thing it lost in 2026: native video generation, after Sora was wound down.

Pricing: powerful, but a maze

In 2026 ChatGPT spans roughly seven tiers — and two of them confusingly share the “Pro” name at different prices. Here’s how it shakes out:

Developers pay separately through the API (GPT-5.5 runs about $5 per million input tokens and $30 per million output tokens) — a subscription includes no API credits, and vice versa. For the vast majority of individual users, Plus at $20 is the sweet spot, and notably it has held that price since 2023.

Pros and cons

Where it falls short

For all its power, ChatGPT in 2026 has rough edges worth knowing before you commit. The biggest is simply complexity: working out which plan and which model you need now takes genuine research, and the naming (5.3 Instant, 5.5, 5.5 Pro) doesn’t help. The free experience has also been squeezed — tight message caps and, since early 2026, contextual ads in the chat.

Reliability remains the perennial issue every large model shares: ChatGPT can still state wrong things with total confidence, and its handling of live sources has been questioned in testing. On the privacy side, the move toward agents and a memory-rich browser means ChatGPT now sees and remembers far more about you, and security researchers have already demonstrated prompt-injection style exploits against the Atlas browser. Finally, losing native video leaves a visible gap against Google’s Veo and others. None of these are dealbreakers for most users, but they’re real.

Scorecard

How ChatGPT scores across the things that actually matter, averaging to 8.8/10:

Verdict

ChatGPT is still the one to beat. Nothing else combines this level of reasoning, this much coding power, and this broad a toolkit in a single product — and the $20 Plus plan remains one of the best-value subscriptions in tech. It loses points not on capability but on coherence: the pricing maze, the ads creeping onto free plans, the familiar reliability caveats, and the gap left by dropping native video.

If you want one AI that does almost everything and sits at the centre of the richest ecosystem, ChatGPT is the safe, strong default. Power users who live in code or research will find real value stepping up to Pro; everyone else should start on Plus. For comparison, see our Claude AI review and our Google Gemini review — the two assistants that push it hardest.

Frequently asked questions

Reviewed June 2026 by AINewsAndUpdates.com. Models, features and pricing change quickly — always confirm current details on the official site before subscribing.