ECOSYSTEM GUIDE · LAST CHECKED NOVEMBER 3, 2026

ChatGPT. The broadest ecosystem with images, voice, and agents

What OpenAI offers right now, which plan reaches which feature, and how to install your context files so you stop re-explaining your business. Prices and features change, so every claim here is checked against OpenAI's own announcements and pricing pages.

Open the AI Tool Dashboard →All ecosystem guides

CURRENT PLANS

What each plan gets you.

Start on the lowest plan that covers the work you do every week. Upgrade when you hit a real limit.

Plan	Price	Best for
Free	$0	Curious users, light usage. Unlimited text chats on GPT-5.6 Luna and a Think button on the web; limits still apply to files and images. Ads in responses.
Go	$8/month	Budget individuals. Unlimited text chats on GPT-5.6 Luna, Think button, ~20 images/day, ad-supported.
Plus	$20/month	Solo professionals and creators. The real tool. GPT-6 Sol in Chat, GPT-6 Astra in ChatGPT Work and Codex, ChatGPT Images 2.5, voice, Deep Research, Canvas, plugins.
Pro	$100, $200, or $500/month	Heavy users. Three monthly plans: Pro 100, Pro 200 (open to new subscriptions again since September 29, 2026), and Pro 500 (launched September 29, 2026, 25 times the Plus allowance, the only one with GPT-6 Astra Ultrafast). All include GPT-6 Astra Pro and maximum Deep Research. Billed monthly only. Since October 28, 2026, each eligible Pro plan can claim matching monthly OpenAI API credits ($100, $200, or $500, promotional, expire each billing cycle).
Business/Team	$25/user/month annually	Small teams 2-150. Everything in Plus, admin controls, data not used for training.
Enterprise	Custom	Larger organizations. SLAs, security, compliance, dedicated support.

CONTEXT INSTALLATION

ChatGPT: Projects + Memory + Plugins (Custom GPTs retire December 11, 2026)

Where your three context files live in ChatGPT, and how to keep them working.

Set it up

Turn on Memory and tell ChatGPT key facts (shortened from who-i-am.md). Create a "Business Brain" Project, paste condensed versions of all three files into the Project instructions, and add the full files to the Project. Do not build new Custom GPTs: they retire December 11, 2026. If you already built master Custom GPTs, select Migrate to plugin under My GPTs when it appears. The instructions become a skill and the knowledge files become reference files.

How it gets used

Memory gives a light, global sense of who you are. Projects are the deeper containers for your context files and instructions. Migrated plugins carry your old GPT instructions as skills you can call by name.

Day to day

Day-to-day, work inside your "Business Brain" Project, not the generic chat. For new initiatives, spin up a new Project with the same three files.

Watch out

Custom GPTs stop working December 11, 2026, and migration does not carry over custom actions, sharing settings, or the selected model. If you pivot, update both Memory and your Project instructions, otherwise ChatGPT pulls old identity details.

TOOLS · 13

Everything inside ChatGPT.

Open a tool to see when to use it, what it needs, and where it falls short.

ChatGPT ChatThe main conversation interface.

What it is: The main conversation interface. Since October 7, 2026, Chat on Plus, Pro, Business, and Enterprise runs GPT-6 Sol with Intelligent UI: answers can include comparisons, diagrams, and small tools you use right in the conversation, and can start while ChatGPT keeps thinking. GPT-6 Astra (September 3, 2026) stays OpenAI's flagship: it powers the Pro option (GPT-6 Astra Pro) on Pro, Business, and Enterprise, and Plus gets it in ChatGPT Work and Codex. Free and Go run on GPT-5.6 Luna by default, with unlimited text chats and a Think button on the web for harder questions (August 2026). OpenAI names models by generation (the number) and tier (Sol, Luna). GPT-6.1 Sol (September 29, 2026) and GPT-6 Luna (September 22, 2026) run in ChatGPT Work and Codex for Plus and up; GPT-6.1 Sol is not in Chat. Free and Go can use GPT-6 Luna in the desktop app. Available in your browser at chatgpt.com, as mobile apps, and as a desktop app. 900M+ weekly active users and 50M+ subscribers, the most-used AI product on the planet. ChatGPT for Clinicians launched as a free verified-clinician tier, and OpenAI is building toward a 'super app' that combines ChatGPT, Codex, and an AI browser. NEW (June 4, 2026): Dreaming, an upgraded memory system that runs a background process synthesizing memory across many conversations. Dreaming auto-updates memories as time passes (e.g., revises 'going to Singapore in July' to 'went to Singapore in July 2026' once the date is past). All memories remain reviewable through a summary page where you can add, edit, or remove what ChatGPT knows about you. Memory Sources (June 2026) adds a control on every model, including Free, that shows which saved memories or past chats were used to personalize a given response, so you can delete or correct anything out of date.
Use it when: You need a quick draft, brainstorm, or one-off task. Great for image generation (ChatGPT Images 2.5 built in), voice conversations, and anything where you want the broadest feature set in one interface.
What you need: Nothing. This is the zero-setup entry point.
Plan required: - Free: $0. GPT-5.6 Luna by default with unlimited text chats (OpenAI applies abuse guardrails) and a Think button on the web. Limits still apply to file uploads, images, and other tools. 2-3 ChatGPT Images 2.5 generations per day. Ads now appearing in responses. - Go: $8/month. GPT-5.6 Luna by default with unlimited text chats and the Think button, ~20 images/day, no Deep Research. Ad-supported. - Plus: $20/month. This is where it becomes a real tool. Chat runs GPT-6 Sol with Intelligent UI (since October 7, 2026). Also brings ChatGPT Work and Codex (running GPT-6 Astra, OpenAI's flagship, plus GPT-6.1 Sol and GPT-6 Luna), Sites for hosted pages, ~3,000 messages/week, ~50 images per 3-hour window, Deep Research, Canvas, and plugins (Custom GPTs retire December 11, 2026). - Pro: three monthly plans. Pro 100 at $100. Pro 200 at $200, open to new subscriptions again since September 29, 2026, with a lower usage allowance for new subscribers. Pro 500 at $500 (launched September 29, 2026), with 25 times the Plus allowance and the only Pro plan with GPT-6 Astra Ultrafast in ChatGPT Work and Codex. All three add GPT-6 Astra Pro. You are buying usage headroom more than features. - Business/Team: $25/user/month annually ($30 monthly, min 2 users). Everything in Plus, centralized billing, no training on your data by default. - Enterprise: Custom pricing. Everything in Business plus SLAs, security/compliance, dedicated support.
Limitation: ChatGPT now has Dreaming (June 4, 2026) which runs a background memory synthesis process and auto-updates memories over time. That solves a real stale-memory problem, but Memory still stores facts, not identity. It remembers your name, preferences, and now historical context. It does not deeply understand your business model, voice, or audience the way Claude Projects with context files do. Output still sounds generic until you build systems around it (plugins, Projects, detailed Custom Instructions).
Closest comparison: Claude Chat (calmer, more editorial writer/researcher vibe). Gemini Chat (Google-integrated, more "search engine with AI" feel).

Custom GPTsRetiring.

What it is: Retiring. On September 11, 2026 OpenAI announced it plans to retire custom GPTs across ChatGPT plans and move them to plugins, which bring reusable instructions, reference files, and connected apps together. Custom GPTs are scheduled to retire on December 11, 2026, when they and their GPT pages become inaccessible. Your past GPT conversations stay accessible. Until then, a Custom GPT is still a specialized version of ChatGPT you configure for one job: instructions, reference docs, and sometimes connected tools.
Use it when: Do not build a new one. If you created a GPT you rely on, finish your edits and publish it (public sharing is not required), select Migrate to plugin under My GPTs when it appears, then test the plugin with a few prompts you use often. If you use someone else's GPT, check its notice and the creator's guidance about a replacement.
What you need: A clear job ("This GPT exists to do X only"). A small set of reference documents. A simple prompt style you want it to follow.
Plan required: Existing GPTs keep working until December 11, 2026. Migration to plugins appears under My GPTs at different times for different accounts. For Enterprise workspaces, OpenAI plans to end creation of new custom GPTs on October 26, 2026 (dates marked subject to change).
Limitation: Every custom GPT becomes inaccessible on December 11, 2026 (February 11, 2027 only for Enterprise workspaces with an approved deferral). Migration copies the latest published instructions and knowledge files into a plugin, but custom actions do not transfer, the selected model does not carry over, sharing settings do not carry over, and the plugin starts private. If you share a GPT with clients or students, each of them needs access to the replacement. The sealed-box problem stays: brand voice across five tasks still means five copies, where one Claude Project shares the same context files across every conversation inside it.
Closest comparison: On ChatGPT itself, plugins are the replacement OpenAI names: the GPT's instructions become a skill, its knowledge files become reference files, and its connected apps come along as apps. Claude Projects plus Skills hold persistent context and instructions. Gemini skills are replacing Gems.

CanvasA side-by-side editing workspace for writing and code.

What it is: A side-by-side editing workspace for writing and code. Instead of generating outputs in the chat window, Canvas opens a dedicated editor where you and ChatGPT collaborate on a document or codebase. You can highlight specific sections for targeted edits without regenerating everything.
Use it when: You are shaping something long and important: a course outline, sales page, flagship presentation. You want to move sections around, ask ChatGPT to rephrase parts, or stitch in new examples without juggling ten windows.
What you need: A draft or at least an outline. Canvas is best for refining and restructuring, not staring at a blank page.
Plan required: Plus, Pro, Team/Business.
Limitation: Easy to accept too many edits and end up with a generic voice. Claude often has the edge for deep editorial nuance. Canvas is also not an artifact you can download and deploy. If you want a deployable web page, you copy the code out manually. Canvas can be slower on huge documents and still not as comfortable as a real word processor for complex formatting.
Closest comparison: Claude Artifacts generate standalone outputs you can download and deploy (HTML/CSS/JS files, interactive tools, documents). Gemini Canvas creates interactive apps and pages from prompts. Key difference: ChatGPT Canvas is an editing workspace. Claude Artifacts are deployable objects. Canvas helps you edit. Artifacts give you files you can host.

Code Interpreter / Advanced Data AnalysisUpload spreadsheets, CSVs, PDFs, images, and ChatGPT analyzes them, runs calculations, creates charts, cleans data, and generates visualizations.

What it is: Upload spreadsheets, CSVs, PDFs, images, and ChatGPT analyzes them, runs calculations, creates charts, cleans data, and generates visualizations. It writes and runs Python code in the background. You don't write code. You ask questions about your data.
Use it when: You have data to analyze: Stripe or ThriveCart sales, client session data, survey exports, content performance metrics. "Show me trends." "Which offer is most profitable?" "What days do people buy most?"
What you need: A reasonably clean file (CSV/Excel from Stripe, ThriveCart, PayPal, email platform). Clarity on what you care about: revenue, retention, refund rates, show-up rates.
Plan required: Plus, Pro, Business, Enterprise.
Limitation: Code Interpreter resets between sessions. If you upload a file and close the chat, the file is gone. You re-upload every session. Claude Projects keep uploaded files accessible across conversations. Also: not a substitute for a proper analytics setup. If your data is messy, it can only guess.
Closest comparison: Claude can analyze files and reason about them, but lacks the full "run Python, generate charts, and keep them in a session" experience. Gemini can analyze files especially tied into Sheets, but advanced session-based analysis is still stronger in ChatGPT.

Image GenerationBuilt-in image generation.

What it is: Built-in image generation. Describe what you want, ChatGPT creates it. The current model is ChatGPT Images 2.5 (September 8, 2026), with sharper detail, more precise editing, and faster generation than the previous Images model. You can start from a template or, on mobile, turn a sketch into an image. Social graphics, course thumbnails, concept mockups, and visual content directly in chat. ChatGPT Images 2.5 supersedes the older DALL-E generation, which has been retired.
Use it when: You need a quick visual without opening a separate design tool. Social media graphics, blog post images, concept mockups, presentation visuals.
What you need: A clear description including style, mood, colors, composition. The more specific the prompt, the better the output.
Plan required: Free gets 2-3 images per day. Go ($8/month) gets ~20/day. Plus ($20/month) gets ~50 images per 3-hour window (~150-200/day). Pro tier ($100/month) get effectively unlimited.
Limitation: ChatGPT Images 2.5 is competent, not exceptional. For brand-consistent visuals with precise text handling, Ideogram often produces better results. Text inside images is still inconsistent. Maintaining exact brand colors across generations is unreliable. If visual brand consistency matters, you will likely still need a dedicated design tool.
Closest comparison: Claude does NOT generate raster images, but Claude Design (launched April 17, 2026) creates designs, prototypes, slides, and one-pagers. For raster: ChatGPT Images 2.5 free tier (2-3 images/day), Gemini free tier (Nano Banana 2 with Personal Intelligence), or Ideogram free. Gemini Nano Banana 2 with Personal Intelligence personalizes outputs using your Google Photos and context.

Voice Mode / Advanced VoiceReal-time voice conversation with ChatGPT.

What it is: Real-time voice conversation with ChatGPT. Not speech-to-text. Full conversational interface where ChatGPT adapts tone, speed, and speaking style based on context. You can interrupt, redirect, and have flowing dialogue. Works with plugins (since September 23, 2026), so you can use your connected apps by voice.
Use it when: Brainstorming while walking, driving, or cooking. Emotionally processing a client situation or offer anxiety. You prefer talking over typing. Great gateway for people intimidated by typing to AI.
What you need: The ChatGPT mobile app or desktop app. Basic privacy rules in mind.
Plan required: Free uses GPT-Live-1.1 Quip with limited access. Go: up to 3 hours a day on GPT-Live-1.1 Quip. Plus: up to 3 hours a day on GPT-Live-1.1. Pro at $100: up to 15 hours. Pro at $200: unlimited (OpenAI's limits as of October 15, 2026). Voice can switch to GPT-5.6 or GPT-6 Astra when it needs to search or reason through harder questions.
Limitation: Voice Mode is impressive for brainstorming but does not produce written output you can use directly. You talk, it talks back, but the transcript is rough. You still need to turn voice conversations into action items or documents separately. It is not a therapist and should not be treated as one.
Closest comparison: Claude has voice mode on mobile (5 voices via ElevenLabs, ~20-30 free conversations/day) but ChatGPT Voice is still seen as more polished for longer conversations. Gemini Live is the direct competitor for real-time voice with camera/screen sharing (strong on Android, free on all tiers).

MemoryChatGPT remembers facts about you across conversations.

What it is: ChatGPT remembers facts about you across conversations. Your name, preferences, business type, tools you use, writing style preferences. You can tell it to remember things and review/edit/delete memories in settings.
Use it when: You want ChatGPT to know recurring context without re-explaining every session. "Remember I run a therapy practice." "Remember my brand colors are lilac and citron."
What you need: Decide what you want remembered (and what you don't). Write a short description of your business, offers, audience, and voice.
Plan required: Plus, Pro, Business, Enterprise.
Limitation: If you let it remember too much, it gets stale or confused, especially when you pivot offers. You need to occasionally clean up memory. Claude's project-based approach can feel more transparent and less "mystical" than global memory. Memory stores isolated facts. It is not a structured understanding of your business the way Claude Projects with context files provide.
Closest comparison: Claude has project-scoped memory and a background memory system that derives info from past conversations. Gemini uses your Google account data to personalize but doesn't have a clear user-controllable memory panel the same way.

ChatGPT ProjectsFolders with memory and tools for a specific goal: launch plan, book, course, client program.

What it is: Folders with memory and tools for a specific goal: launch plan, book, course, client program. You keep chats, files, instructions, and tools together so ChatGPT can see everything when working on that project, instead of each chat being isolated.
Use it when: You are working on a single big thing (book, certification program, full rebrand) and want one place where all drafts, research, and decisions live.
What you need: Decide the project boundary: "This project is 'New group program,' not my whole business." Gather relevant files and upload them.
Plan required: Plus, Pro, Team/Business.
Limitation: You still need to be the project manager. ChatGPT won't automatically organize everything. Claude's Projects can feel more transparent and less feature-heavy, which can be easier mentally.
Closest comparison: Claude Projects are very similar conceptually and often praised for clarity. Gemini has projects via AI Studio and Google Docs/Drive, but more developer-oriented.

CodexAn autonomous agent that started as a coding tool and expanded into a full desktop agent (major update April 16, 2026).

What it is: An autonomous agent that started as a coding tool and expanded into a full desktop agent (major update April 16, 2026). Codex now works at the operating system level on Mac with its own background cursor. It clicks, types, and navigates apps in parallel while you work on something else. This is the key difference from Claude Cowork (which takes over your actual screen). Codex runs in a sandboxed virtual workspace so multiple agents can operate apps simultaneously without interrupting you. Also includes: in-app browser, image generation, persistent memory, scheduled/long-running automations, and 90+ plugins (Atlassian, GitLab, Microsoft Suite, Remotion, and more). NEW (June 2, 2026): six role-specific plugins bundling 62 apps and 110 skills (Corporate Finance, Private Equity Investing, Marketing Strategy, Strategy Consulting, Legal, plus more coming). Sites lets Codex create and share interactive hosted websites and apps directly (dashboards, planners, project boards, galleries, lightweight tools) shareable via URL inside your workspace. Per OpenAI, non-developers now make up ~20% of Codex users and are growing more than 3x as fast as developers. Available as a desktop app (Mac/Windows), CLI, web, and IDE extensions.
Use it when: You or a contractor are building software and want to delegate coding tasks. Also now: any repetitive computer task you can describe in plain language, as long as you are on Mac. Feature development, bug fixes, PR reviews, data entry, report generation across apps.
What you need: For coding: a GitHub repository connected. For computer use: Mac with the Codex desktop app. Basic review ability (you need to check what Codex produces). Comfort giving an agent limited control over your apps.
Plan required: Plus ($20/month) and above, running GPT-6.1 Sol (September 29, 2026), GPT-6.1 Terra (October 26, 2026), and GPT-6 Luna. Pro comes in three monthly plans ($100, $200, $500) for more usage headroom, and only Pro 500 includes GPT-6 Astra Ultrafast. Codex Cloud (September 29, 2026) runs coding tasks in reusable cloud environments from desktop, web, or mobile, and tasks continue while your computer is asleep. Computer use (background cursor) is Mac-only and not available in EEA, UK, or Switzerland.
Limitation: Computer use is Mac-only. Windows users have the app but not cursor-level control yet. Parallel background operation is powerful but raises security concerns. Help Net Security recommends a dedicated macOS user account for Codex to avoid it seeing sensitive data across all your apps. For non-technical founders building their first tool, Claude Code is still more accessible because it works locally with plain-language instructions and does not require understanding pull requests.
Closest comparison: Claude Code plus Claude's task handoff (Cowork merged into Claude on September 16, 2026). Claude Code handles coding, and tasks handle multi-step work that keeps running after you close your laptop. Gemini Code Assist + Jules (catching up, best for Google Cloud). The Codex advantage is parallel background operation. The Claude advantage is writing quality and instruction precision (especially on Opus 5.5, which Anthropic says performs at the level of Fable 5.1 on most work). Over 4 million active developers use Codex as of April 2026, with a new student free tier.

Video GenerationChatGPT has had video generation through a product called Sora.

What it is: ChatGPT has had video generation through a product called Sora. Sora is shut down: OpenAI discontinued the Sora web and app experiences on April 26, 2026, and set the API to end September 24, 2026. OpenAI is reallocating Sora's resources toward ChatGPT as the core product. Video generation capabilities are expected to continue inside ChatGPT through newer integrated tools, but not under the Sora brand.
Use it when: You want short promotional clips for launch pages, social posts, or course intros. The landscape is actively shifting. Do not build workflows around Sora specifically.
What you need: Tight storyboard ideas. Clarity on where the video will be used.
Plan required: Limited access on Plus, Pro, Business, Enterprise with separate usage caps.
Limitation: Sora as a distinct product is gone. OpenAI's help center points Sora creators to sora.chatgpt.com/sunset to export what they made. Future video generation will live inside ChatGPT directly under a different name, or through Gemini's Veo. Quality is inconsistent. Not yet reliable enough to replace real video production for professional content.
Closest comparison: Gemini has Veo 3/3.1 for video generation (with sound, industry leading). Claude does not generate video.

Deep ResearchAn autonomous research agent that plans, searches the web, reads sources, synthesizes findings, and produces long cited reports.

What it is: An autonomous research agent that plans, searches the web, reads sources, synthesizes findings, and produces long cited reports. Built for deep research projects, not quick answers.
Use it when: Market scan, regulatory research, curriculum topic deep-dive, competitive analysis. Any question where you need depth and evidence.
What you need
Plan required: Plus, Pro, Team/Business.
Limitation: Slow (several minutes per query). Uses significant capacity from your message allowance. Can miss key sources. Claude is often more transparent in written research. Gemini is strongest when you care about Google-indexed sources.
Closest comparison: Claude Deep Research (often wins on clarity and tone). Gemini Deep Research (can access your Gmail, Drive, and Chat data with permissions).

OperatorChatGPT's browser-based agent for web tasks.

What it is: ChatGPT's browser-based agent for web tasks. Navigates websites, fills forms, clicks buttons. Note: the April 16 Codex update now handles desktop-level computer use (clicking apps, navigating software) with its own background cursor on Mac. Operator focuses on web tasks. Codex handles desktop tasks. They are complementary, not the same product.
Use it when: You have repetitive computer tasks: uploading content to a platform, copying data between tools, downloading reports. You want to automate weekly client report creation across multiple apps.
What you need: Clear step-by-step instructions for the task. Comfort giving an AI agent limited control over your workspace.
Plan required: Pro, Business, Enterprise (Operator). Codex computer use: Plus and above, Mac-only.
Limitation: Still early and brittle. Interface changes or pop-ups can break flows. For higher-risk tasks (money, compliance), keep a human in the loop. Claude and Gemini currently lag here too, but that also means fewer surprises from agents doing the wrong thing.
Closest comparison: Claude Cowork (desktop automation, takes over your screen). The Codex desktop agent (parallel background operation). Gemini is piloting similar behaviors in Workspace but not at the same breadth.

Other ChatGPT CapabilitiesWeb Browsing/Search: Available on Go and above.

What it is: Web Browsing/Search: Available on Go and above. Fetches current information. Gemini may be stronger for pure search tasks (it is literally Google). Claude tends to be more careful but less integrated with real-time search. File Uploads and Analysis: PDFs, Word docs, spreadsheets, PowerPoints, audio/video. Plus/Pro/Business get higher limits. Claude usually performs better on very long or complex documents. Gemini excels when docs already live in Drive. Plugins and connected apps: Google Drive, Dropbox, Outlook, Gmail, GitHub, HubSpot, and more, on all ChatGPT plans. Since September 17, 2026 you can connect more than one account (personal and work) to a plugin. Gemini may be more seamless if your whole life is in Google Workspace. Thinking options: pick a Thinking option in the model picker when you want ChatGPT to reason longer. Since September 14, 2026, Plus and Pro no longer switch from Instant to Thinking automatically.
Use it when
What you need
Plan required
Limitation
Closest comparison

SIDE BY SIDE

How ChatGPT compares.

Supported does not mean identical. Use this when you are choosing where a workflow should live.

Capability	ChatGPT	Claude	Gemini
Writing quality (clarity, nuance, ethics)	Competitive. Excellent but more "performer" and can be too confident.	Winner. Best for thoughtful, grounded, long-form writing and sensitive topics.	Competitive. Solid but more "Google-corporate" in tone.
Brand voice consistency	Competitive. Memory + plugins are powerful but can get fragmented. Custom GPTs retire December 11, 2026.	Winner. Projects + Skills make it easy to keep one consistent tone.	Weaker. Tooling less focused on brand voice.
Image generation	Winner. ChatGPT Images 2.5 integrated directly in chat (supersedes the older DALL-E generation).	Missing. Understands images only. Claude users can use ChatGPT free, Gemini free, Microsoft Designer free, or Ideogram free.	Competitive. Nano Banana and Imagen 4 strong, free tier generous, especially for Workspace.
Voice conversation	Winner. ChatGPT Voice is the most natural "talk to AI" experience, with daily limits set by plan (Plus gets up to 3 hours on GPT-Live-1.1).	Competitive. Voice mode on mobile, ~20-30 free conversations/day, English-only.	Competitive. Gemini Live free on all tiers, strong on Android, camera and screen sharing.
Video generation	Weaker. Sora shut down April 26, 2026 (API end date September 24, 2026). No named replacement yet.	Missing. No native video.	Winner. Gemini Omni creates and edits video from any input. Veo 3.1 still leads on cinematic, sound-on clips, free tier includes limited Veo 3.1 Lite via Flow.
Coding agent	Competitive. Codex + GPT-6.1 Sol (September 29, 2026) and GPT-6 Astra, with background computer use and 90+ plugins. Strongest ecosystem breadth.	Winner (developer preference). Claude Code + Opus 5.5 (launched September 22, 2026; Anthropic says it performs at the level of Fable 5.1 on most work at 40% lower cost than Opus 5). Routines add scheduled automation. Dynamic Workflows on Max+.	Competitive. Best for Google Cloud teams.
Deployable artifacts and design	Competitive. Canvas + code blocks for editing. No design-system-aware visual generation.	Winner. Artifacts for deployable tools. Claude Design (April 17) adds branded prototypes and visual documents with design system ingestion. Claude Slides and Claude Docs (September 16, 2026, on every plan since October 8, 2026) add decks you can download as PowerPoint or PDF and documents you write with Claude.	Competitive. Canvas for apps/pages. No equivalent to Claude Design.
Productivity suite integration	Competitive. ChatGPT add-ins for Excel, Word, and PowerPoint plus Google Sheets on every plan, Free included.	Competitive. Claude for Microsoft 365 and Claude for Microsoft Outlook are included on Pro and up, plus MCP connectors for other tools.	Winner. Deeply embedded in Gmail, Docs, Sheets, Slides, Meet.
Persistent context / projects	Competitive. Projects + Memory + plugins work but can scatter. Custom GPTs retire December 11, 2026.	Winner. Projects + Skills give clear, scoped "brains" per initiative.	Competitive. Skills (replacing Gems) + Notebooks + Gemini Notebook powerful but more fragmented.
Source-grounded research	Competitive. Deep Research powerful.	Competitive. Deep Research strong.	Winner (situational). Gemini Notebook is built around your own sources: answers come with citations you can check.
Desktop automation	Winner (new). Codex (April 16 update) runs with its own background cursor in parallel. Multiple agents, no screen takeover. Mac-only.	Competitive. Claude takes on multi-step tasks from any conversation and keeps working after you close your laptop (Cowork merged into Claude September 16, 2026, Pro and Max first).	Weaker. Desktop apps on Mac and Windows (one-click window sharing on Mac) but no computer use. Workspace actions only.
Scheduled automation	Competitive. Codex now supports scheduled/long-running automations via desktop app.	Competitive. Claude Code Routines (April 14, 2026) run prompts on cron, webhook, or GitHub triggers. Pro: 5/day, Max: 15/day. Research preview.	Winner (still). Gemini Scheduled Actions are simplest for non-technical users. Routines and Codex are more powerful but more technical.
Free tier generosity	Competitive. Unlimited text chats on GPT-5.6 Luna with a Think button since August 2026. Ads in responses.	Competitive. Solid but usage caps.	Winner. Strong free tier, especially with Google integration.
Best value at $20/month	Winner for generalists. Plus is most well-rounded (text, images, voice, code).	Winner for writers. Pro is unbeatable for thinking and writing.	Winner for Google users. AI Pro is unbeatable for Gmail/Docs/Sheets plus 5TB storage.