ChatGPT
OpenAI's flagship AI assistant for chat, coding, image and video generation, and agentic tasks.
Overview
ChatGPT is OpenAI's flagship AI assistant for work, code, and creativity
ChatGPT is the most widely used AI assistant in the world, built on OpenAI's GPT-5 family of models. Originally launched in November 2022 as a research preview, ChatGPT has grown into a multi-modal product that handles writing, coding, analysis, agentic tasks, image generation, and video generation across web, desktop, and mobile. As of April 2026, the default model is GPT-5.5 (launched April 23, 2026), with GPT-5.5 Pro reserved for Pro, Business, and Enterprise tiers.
Key Features (April 2026)
- GPT-5 model family — GPT-5.5 default, GPT-5.5 Pro on higher tiers; lighter GPT-5.3 variants for free use.
- Codex Agent — autonomous coding agent that writes, reviews, and ships code from inside ChatGPT or via CLI.
- Agent Mode — multi-step task automation (browse, click, fill forms, execute workflows).
- Deep Research — long-running, citation-backed research reports (10/mo on Plus).
- Canvas — dedicated long-form workspace for writing and code (Plus and above).
- Sora video — text-to-video and image-to-video generation; 50 videos/mo on Plus (note: the standalone Sora app and API are being deprecated in 2026; in-ChatGPT video gen continues).
- Image generation — native image creation with strong prompt adherence and editing.
- Advanced Voice Mode — natural conversational voice with translation; available to free users with time limits.
- Projects, Tasks, Memory — persistent context, scheduled jobs, and per-project organization.
- Custom GPTs / GPT Store — build and publish task-specific assistants.
Pricing (April 2026)
- Free — $0. GPT-5.3 with tight limits, ads (in US).
- Go — $8/mo. Higher message volume, no advanced reasoning, no Sora/Codex/Agent Mode/Deep Research.
- Plus — $20/mo. Full GPT-5.5 + Codex, Agent Mode, Deep Research (10 runs/mo), Sora video, Canvas.
- Pro — $100/mo or $200/mo. Heavy use limits, GPT-5.5 Pro, expanded tool access.
- Business — $25/user/mo. Shared workspaces, SSO, admin tools, usage visibility.
- Enterprise — custom. Compliance, data residency, virtually unlimited messages.
Best For
- Knowledge workers — drafting, research, summarization, and analysis at scale.
- Software developers — Codex Agent for writing, reviewing, and shipping code from chat or CLI.
- Marketers and content teams — long-form drafts, campaign briefs, and image/video assets in one tool.
- Operators and analysts — Agent Mode for multi-step workflows; Tasks for scheduled automations.
- Solo founders and consultants — Plus at $20/mo unlocks the full toolbelt for daily work.
ChatGPT vs. Claude vs. Gemini
- ChatGPT — broadest product surface (Codex, Sora, Agent Mode, voice, image, custom GPTs); largest user base; deepest plugin ecosystem.
- Claude — strongest at long-context reasoning and coding judgment; safety-first; 1M-token context on Opus 4.7.
- Gemini — tightest Google Workspace integration; competitive free tier with large context.
FAQ
Is ChatGPT free? Yes. The Free tier runs GPT-5.3 with tight message and upload limits and (in the US) ads. Go ($8/mo) lifts limits but excludes advanced reasoning and Sora/Codex/Agent Mode.
What's GPT-5.5? The default flagship model from April 23, 2026, replacing GPT-5.4. GPT-5.5 Pro is a higher-accuracy variant restricted to Pro, Business, and Enterprise tiers.
Can ChatGPT write code for me? Yes. Codex Agent is included on Plus, Pro, Business, and Enterprise. It runs inside ChatGPT or as a CLI tool that can read your repo, run tests, and ship pull requests.
Can ChatGPT generate video? Yes. Plus users get 50 Sora videos per month inside ChatGPT. Note: the standalone Sora app shut down April 26, 2026, and the public Sora API is being discontinued September 24, 2026 — but video generation inside ChatGPT continues.
What is Agent Mode? A built-in agent that completes multi-step tasks for you — opens a browser, clicks, fills forms, and executes workflows on your behalf. Available on Plus and higher.
Does ChatGPT remember me between sessions? Yes. Memory is on by default and stores facts across conversations. You can clear or disable it in settings.
Related
Looking for more options? Browse the Productivity directory or read our best AI productivity tools listicle. ChatGPT has a Wikipedia entry and is tracked on Crunchbase.
Why Use ChatGPT
FAQ
Editorial Review
Our take on ChatGPT.
Still the default AI subscription, three years running. GPT-5.5 is the most-shipped frontier model and nothing rivals the surface area — chat, voice, images, Codex agents, search, and memory on one login. Rivals now win specific jobs; that is how you should buy.
What works
- Widest surface from one login — GPT-5.5 chat, voice, vision, images, Codex agents, Search, and memory; the alternative is still three or four separate bills.
- Codex is the most accessible agentic-coding on-ramp on the market — included with 400K context from the $8 Go tier up.
- GPT-5.5 wins agentic terminal execution outright (Terminal-Bench 78.2 vs Opus 4.8's 74.6) — best-in-market computer use.
- ChatGPT Search delivers sourced answers inline and has absorbed the casual research use case.
- Cross-session memory compounds for daily users — it knows your team, style, and codebase by week two.
What doesn't
- Video is gone — Sora shut down on economics in March 2026; rivals bundle nothing comparable either, but the old "one bill covers video" pitch is dead.
- For code review and high-stakes analysis, Opus 4.8 is measurably more likely to catch what this model lets slide.
- Quota ceilings on Thinking/Pro models are still not telegraphed — heavy days end in quiet downshifts.
- Image generation trails Midjourney v7 on aesthetics.
- Pro-tier value moves without much notice — the 10× Codex allowance settled to 5× after May 31.
ChatGPT holds the default-subscription spot for the third year running, but 2026 changed what you're buying. GPT-5.5 (April 2026) is the headline — noticeably better at carrying multi-step work end-to-end — and the Codex agent is now the real product story. The model is not uniformly the best: Claude Opus 4.8 wins on coding depth and on flagging problems it could quietly let slide, and Gemini 3.5 Flash is roughly 70% cheaper and several times faster for volume work. The bundle is still the strongest single buy for a generalist.
What you actually pay for
The ladder runs Free, Go at $8, Plus at $20, Pro at $100, and Pro at $200, with Business seats at $25–30. Plus gets GPT-5.5 Thinking; the Pro tiers add GPT-5.5 Pro for higher-accuracy work. Even the $8 Go tier now includes GPT-5.5 inside Codex with a 400K context window — the cheapest agentic-coding entry point from any frontier lab. On top of the models: voice mode, vision input, image generation, Code Interpreter, custom GPTs, cross-session memory, and ChatGPT Search, which has quietly become the default research surface for most subscribers. The competitive answer to that list is still several separate bills.
One fine-print note: the $100 Pro tier launched with a 10× Codex allowance that settled to 5× Plus after May 31. Quotas on this ladder move; read the current numbers before upgrading.
The Sora lesson
The bundle lost a modality this year. OpenAI shut Sora down in March 2026 — the app went dark April 26 and the API follows in September — after costs near $1M a day met a user base that had fallen under 500K. We track the shutdown in our AI Graveyard: even OpenAI retires products that don't pay for their compute. If video generation is your primary need, buy Runway, Google Veo, or Kling directly; nothing in ChatGPT covers it today.
Where it falls behind
Three durable complaints. First, for code review and high-stakes analysis, Opus 4.8 is the safer pick — it resolves 88.6% of SWE-bench Verified issues and is measurably more likely to flag a flaw than agree with your framing. GPT-5.5 still wins agentic terminal work (78.2 vs 74.6 on Terminal-Bench), so the coding crown is genuinely split. Second, quota opacity persists: Thinking and Pro ceilings are not telegraphed, and heavy days still end in quiet downshifts. Third, image generation is competent but visibly behind Midjourney v7 on composition and lighting.
Sycophancy — the model agreeing with whatever framing you hand it — is improved in GPT-5.5 but not gone. We still prompt against it on anything with stakes: "argue the opposite case, then compare."
Who should buy
Plus at $20 is the right answer for a generalist who wants one subscription to cover most knowledge work, for a non-technical team adopting AI for the first time, and for anyone who touches voice, images, and chat in the same week. Go at $8 is the right answer for a coder who mainly wants Codex. It is not the right answer for dedicated coding depth (Claude Code or Cursor), professional image work (Midjourney), or video (Runway or Veo).
The honest comparison
The 2026 pattern among teams we track is routing, not loyalty: ChatGPT for breadth and agentic execution, Claude for the work you need to be right, Gemini for volume and speed. If you buy one this quarter, ChatGPT covers the most surface; see our maintained head-to-head at /compare/chatgpt-openai-vs-claude.
Re-check triggers
We re-rate when OpenAI ships a video successor (a world-model project is rumored for the API), when Pro-tier quotas move again, or when the next frontier model resets latency and pricing.
User Reviews
Flávia Morais
9/9/2023
Amazing tool!
Similar Tools
