Updated 30 October 2026: Since 7 October GPT-6, powered by GPT-6 Sol, has been the default in ChatGPT’s Chat on Plus and above, while Free and Go stay on GPT-5.6 Luna. OpenAI also cancelled the lower Pro 200 allowance for new subscribers on 29 October, so we have updated the models, plans and sources; our verdict is unchanged.
Most people who want an AI assistant end up choosing between Claude and ChatGPT, and this month both got a model refresh: Anthropic shipped Claude Opus 5.5 on 22 September 2026, and OpenAI put GPT-6 Sol and Luna into ChatGPT Work and Codex the same day. Anthropic then followed with Claude Sonnet 5.5 on 28 September, and a month later it filed to list on the stock market, which we cover in our Anthropic IPO explainer. Which one fits you depends less on benchmark tables than on the shape of your work, so we look at what each side gives you for writing, coding and everyday use, what the plans cost, how much context the apps actually let you use, and who trains on your chats unless you tell them not to. Every figure comes from Anthropic’s and OpenAI’s own pages, checked on 28 September 2026, with OpenAI’s rechecked on 30 September after its 29 September announcements and Anthropic’s model pages rechecked the same day for Sonnet 5.5.
- Same headline price, different shape. Claude Pro and ChatGPT Plus are both $20 a month. Below that, ChatGPT has an $8 Go tier that may show ads, while Claude has nothing between Free and Pro. Pay annually and Claude Pro drops to $17 a month.
- Claude gives you far more context in the app. Paid Claude plans give Fable 5.1 and Opus 5.5 the full 1M-token window, and the new Sonnet 5.5 is a 1M-token model too. ChatGPT Plus gets 54K tokens for Instant and 256K for reasoning; Pro raises those to 128K and 400K.
- ChatGPT does more things. It generates images, its Live voice lets you talk over it, and it has scheduled tasks, custom GPTs and a built-in browser. Claude doesn’t generate photos or illustrations at all.
- The privacy defaults are different. A personal ChatGPT account is used for training until you switch off “Improve the model for everyone”. Claude makes you choose at sign-up, and that choice also sets how long your data is kept: five years if you allow training, 30 days if you don’t.
The models behind each app
Claude and ChatGPT are the apps; the thing that actually answers you is a model, and both companies now run three of them. Anthropic’s current line-up is Claude Fable 5.1 (released 1 September 2026), Claude Opus 5.5 (22 September) and Claude Sonnet 5.5 (28 September), which replaced the Sonnet 5 of 30 June. OpenAI’s newest are GPT-6 Astra (3 September), GPT-6 Sol and Luna (22 September) and GPT-6.1 Sol (29 September), and since 7 October GPT-6 has been the default in the ChatGPT app on Plus and above, while Free and Go still run GPT-5.6 Luna. Here are the models you’ll meet in the apps, with the API prices each company publishes.
| Model | API price per 1M tokens (in / out) | Context / max output | Knowledge cutoff | Where you get it in the app |
|---|---|---|---|---|
 Claude Fable 5.11 Sep 2026 | $10 / $50 | 1M / 128K | Jun 2026 | Max, 50 % of weekly limit Pro with usage credits |
| Claude Opus 5.522 Sep 2026 | $4 / $20 | 1M / 128K | Jun 2026 | Pro and Max |
| Claude Sonnet 5.528 Sep 2026 | $2 / $10 | 1M / 128K | Jun 2026 | Pro and Max Free runs “Sonnet”, version not stated |
 GPT-6 Astra3 Sep 2026 | $10 / $50 | 1.05M / 128K | Apr 2026 | Pro, as “GPT-6 Pro” Plus in Work and Codex |
| GPT-6.1 Sol29 Sep 2026 | $2 / $10 | Not yet confirmed | Not yet confirmed | Work and Codex only Plus and Pro; not yet in Chat |
| GPT-6 Sol22 Sep 2026 | $2 / $10 | 1.05M / 128K | Apr 2026 | Chat default on paid plans Plus and Pro, plus Work and Codex |
| GPT-5.6 Sol / Lunastill in Chat | Not sold separately by plan | See plan limits below | Not published on the plan page | Luna on Free and Go Sol still selectable on paid plans |
Two things stand out if you use the apps rather than the API. The price ladders now line up almost exactly, with $10 and $50 at the top and $2 and $10 in the middle. And since 7 October a ChatGPT subscriber talks to GPT-6 in a normal chat. OpenAI’s help centre says GPT-6 Sol now powers Instant, Medium, High and Extra High on paid plans, that GPT-5.6 Sol can still be picked from the model menu, and that Free and Go users still get GPT-5.6 Luna. GPT-6.1 Sol, which arrived on 29 September, hasn’t reached Chat yet: OpenAI says it “is not yet available in Chat” and is only in ChatGPT Work and Codex. In Claude’s app, a Pro subscriber can pick Opus 5.5 or Sonnet 5.5 from the model menu on the day they’re released.
Plans and prices
Both companies sell a free tier, a $20 tier and something at $100 or more, so the differences are at the edges. ChatGPT has an $8 Go plan between Free and Plus, which OpenAI’s pricing page says “may include ads”. Claude has no ad-supported tier and nothing under $20, but the annual Pro price works out at $17 a month. At the top, ChatGPT Pro now comes in three monthly-only tiers: Pro 100 at $100, Pro 200 at $200 and, since 29 September, Pro 500 at $500. Pro 200 is open to new subscribers again, and on 29 October OpenAI cancelled the lower usage allowance it had announced for new Pro 200 subscriptions. Pro 500 has OpenAI’s highest allowance, 25 times Plus, and is the only Pro tier that includes Ultrafast, OpenAI’s faster premium speed tier. Claude Max is $100 for five times Pro’s usage or $200 for twenty times. If Gemini is on your shortlist too, we compare all three plans in Claude, ChatGPT or Gemini.
Before you pay $100 for ChatGPT Pro to get GPT-6 Astra, read this line from OpenAI’s help centre: “A Pro subscription does not include unlimited use of the GPT-6 Pro model.” On Pro 100 you get 50 messages a week, shared with GPT-5.6 Sol Pro; on Pro 200 it’s 200 a week. OpenAI had planned a lower overall usage allowance for new Pro 200 subscriptions, but cancelled it on 29 October 2026, so Pro 200 keeps the same limit as before. Claude has a similar gate on Fable 5.1. It isn’t on the Free plan, Pro users pay for it with usage credits, and Max users can spend up to half their weekly limit on it.
Context windows
Context is how much a model can hold in mind at once: your documents, the conversation so far and its own reply. This is where the two apps differ most, and it’s easy to miss, because both companies sell developers a window of about a million tokens. The question is how much of that the consumer app lets you use.
Anthropic’s help centre says that on paid Claude plans, Fable 5.1, Opus 5.5 and Sonnet 5 all get the full 1M tokens in chat, which Anthropic puts at roughly 555,000 words. When we checked that page on 28 September it named Sonnet 5; its successor, Sonnet 5.5, also has a 1M-token window. OpenAI’s pricing page gives plan-by-plan figures for ChatGPT: the “GPT Instant total context window” is 27K tokens on Free, 54K on Go and Plus, and 128K on Pro, or about 12, 40 and 250 pages of text by OpenAI’s own reckoning. Reasoning modes get 256K on Go and Plus and 400K on Pro, roughly 320 and 680 pages. So at $20, Claude’s window is about four times ChatGPT’s reasoning window and nearly twenty times its Instant window.
Writing, coding and everyday use
We haven’t run a blind head-to-head for this guide, and we’d be wary of anyone who claims a clean winner. Both companies publish benchmark tables that put their own model on top. OpenAI’s GPT-6 Astra page, for instance, shows Astra ahead of Claude Fable 5.1 on Terminal-Bench 4.0 and behind it on Humanity’s Last Exam. What we can do is describe what each app is built for.
Writing. Both are capable drafting partners with memory that carries over between chats. Claude’s advantages are the context window and its Projects, which hold instructions and reference files and, on paid plans, switch to retrieval as the knowledge base grows. There’s also Claude Design, in research preview for Pro and above, for slides, one-pagers and prototypes that export to PDF, PPTX or Canva. ChatGPT’s advantage is breadth. It will generate the images for your document (Claude won’t), it has had a ChatGPT for Word add-in on all plans since 17 September, and its deep research mode is more generous on Plus and Pro than on Free.
Coding. On both sides, this is where the newest models show up first. Claude Pro includes Claude Code, Anthropic’s terminal agent, with a 1M-token window on the newer models. ChatGPT Plus includes Codex, and Codex is where OpenAI puts GPT-6 Astra, Sol and Luna, and now GPT-6.1 Sol, before they reach ordinary chat. Anthropic pitches Opus 5.5 “for long-running agentic coding and knowledge work” and claims the new Sonnet 5.5 scores 70.6 percent on Terminal-Bench 4.0 against 66.4 percent for Opus 5.5; OpenAI says GPT-6 Sol matches Claude Fable 5.1 on its FrontierCode benchmark “at much lower cost”, and claims GPT-6.1 Sol scores 2.2 points above Opus 5.5 at medium effort on AutomationBench at roughly a third of the cost. Treat all of these as marketing until you’ve run your own repository through each.
Everyday use. ChatGPT is the more complete consumer product. Its Live voice mode listens and speaks at the same time, and in Advanced mode it can see through your phone camera or share your screen. It generates and edits images (Images 2.5 shipped on 8 September), runs scheduled tasks from the Go plan up, and has a built-in browser on every plan. Since 29 September OpenAI has also been rolling out dots, always-on agents powered by GPT-6 Astra with their own cloud computer, to eligible Pro and Business Premium users, though Pro access excludes the European Economic Area, Switzerland and the UK at launch. Claude’s voice mode is still labelled beta. It works on every plan and on web, desktop and mobile, and can use web search and your connected tools while you talk, but there’s no camera mode and no image generator. Anthropic’s help centre says plainly that Claude “doesn’t generate photos or illustrations”, only diagrams and charts as HTML or SVG.
+Reasons to pick Claude
- 1M-token context on every paid plan, against 54K / 256K on ChatGPT Plus
- The newest model is in the chat menu on launch day
- Files up to 500 MB and PDFs up to 1,000 pages per upload
- You choose the training setting at sign-up, and opting out shortens retention to 30 days
- $17 a month on the annual plan
+Reasons to pick ChatGPT
- Image generation and editing, limited on Free and unlimited on Pro
- Live voice with interruptions; video and screen sharing on mobile
- $8 Go tier if you only need a little more than Free
- Scheduled tasks, custom GPTs, Sites, a built-in browser
- Shared projects on every plan, including Free
Memory, projects and files
Both remember you between conversations and let you see and delete what they’ve stored. Claude’s memory is on by default for Free, Pro and Max. It stores what it learns as a list of named topics you can edit in Settings, keeps a separate memory per project, and can import memory exported from another assistant or export its own. ChatGPT’s “improved memory” keeps a running summary you correct in place and shows a Sources line under answers that drew on your history; the pricing page calls it “Limited” on Free and “Expanded” on Plus and Pro. Incognito chats in Claude and temporary chats in ChatGPT both stay out of memory.
Who trains on your chats
If you read one section, read this one. For personal ChatGPT accounts (Free, Go, Plus and Pro), OpenAI’s help centre says “we may use your content to train our models”, and the pricing page lists training for all four plans as “Opt-out available”. To opt out, turn off “Improve the model for everyone” under Settings, Data controls, or file a request in OpenAI’s privacy portal. Temporary chats aren’t used for training. Business, Enterprise, Edu and API content isn’t used by default.
Anthropic has done it differently since its consumer terms changed on 28 August 2025: you pick the setting during sign-up and can change it any time under Settings, Privacy, “Help improve our AI models”. The choice has a second effect. Allow training and Anthropic may keep de-identified data for up to five years; decline and it’s 30 days. Incognito chats are never used, whichever you chose. Team, Enterprise and API use is excluded from consumer training entirely.
Where each one runs
Both run in a browser, on iOS and Android, and as desktop apps for macOS and Windows. Claude’s download page also lists Linux, Windows on Arm and ChromeOS, plus extensions for Chrome, Slack, Excel, PowerPoint, Word and Outlook (beta), and Claude Code in VS Code, JetBrains and the terminal. The Chrome extension, which lets Claude click and type on websites for you, is for paid plans only. ChatGPT has a desktop app for macOS and Windows, a Chrome extension, and a Microsoft add-in for Word, Excel and PowerPoint that OpenAI says works on all plans including Free. So if you want the assistant driving your browser, Claude needs Pro, while ChatGPT lists a built-in browser on every plan.
Verdict
Pick Claude if your work is long: big documents, whole codebases, research that runs over weeks. The 1M-token window at $20, Opus 5.5 in the menu on launch day, and a privacy model that shortens retention when you opt out are concrete advantages, and none of them depends on a benchmark. Pick ChatGPT if you want one app that also makes images, talks back naturally, runs tasks on a schedule and works inside your browser, or if $8 a month is the budget. On raw model quality the lead changes hands every few weeks, and both companies’ own tables show them trading wins.
Still can’t decide? Both free tiers are good enough to settle it in a week. Give each the same real task and see which one you’re still opening on day six. If neither fits, our guide to ChatGPT alternatives covers other assistants and what each does better.
Frequently asked questions
Is Claude better than ChatGPT?
Not across the board, no. Claude gives you a much larger context window in the app and puts its newest models in the chat menu on launch day. ChatGPT does more things: image generation, live voice with video, scheduled tasks, a built-in browser. Both companies publish benchmark tables where their own model wins some rows and loses others, so be careful with any single “winner” claim.
Which is cheaper, Claude Pro or ChatGPT Plus?
They cost the same, $20 a month. Claude Pro drops to $17 a month if you pay annually. ChatGPT has a cheaper $8 Go plan, which OpenAI says may include ads, and both companies’ top consumer tiers start at $100 a month.
Can Claude generate images?
No. Anthropic’s help centre says Claude doesn’t generate photos or illustrations. It can build diagrams, charts and interactive visuals as HTML or SVG, and Claude Design (research preview, Pro and above) produces slides and layouts. ChatGPT generates and edits images on every plan, with limits on Free.
Does ChatGPT train on my conversations?
On personal plans, yes, until you turn off “Improve the model for everyone” in Settings under Data controls. Temporary chats are excluded. Claude asks you to choose at sign-up; decline and your data is kept for 30 days instead of up to five years. On both services, rating a reply can send that conversation for training whatever the setting.
Which one is better for coding?
We can’t say without a proper test, and we haven’t run one. Both bundle a coding agent with the $20 plan: Claude Code with Claude Pro, Codex with ChatGPT Plus. OpenAI’s newest GPT-6 models reach Codex before ordinary chat, and Anthropic positions Opus 5.5 for long-running agentic coding. Run your own repository through both free tiers and see.
Sources and method
All prices, model names, release dates, context windows, file limits and privacy defaults were read from Anthropic’s and OpenAI’s own pages on 28 September 2026; OpenAI’s pages were rechecked on 30 September 2026 after its 29 September announcements, and Anthropic’s model pages were rechecked on 30 September 2026 after the Claude Sonnet 5.5 release. The benchmark comparisons we quote are the vendors’ claims from their launch posts, not our measurements; we haven’t run a head-to-head test for this guide. Where the two companies describe a feature differently, we quote their wording rather than reconcile it.
- Anthropic, Models overview: model names, API prices, context, output, cutoffs
- Anthropic, Claude Fable 5.1 model page: release date 1 September 2026
- Anthropic, Introducing Claude Fable 5.1 and Mythos 5.1: announcement figure
- Claude, Pricing: Free, Pro, Max and Team plans and features
- Claude Help Center, What is the Max plan? and Claude Fable models on your plan
- Claude Help Center, context window on paid plans
- Claude Help Center, file upload limits
- Claude Help Center, memory, voice mode, projects, images, Claude in Chrome
- Anthropic, Introducing Claude Design and Claude, Download
- Anthropic, Updates to Consumer Terms and Privacy Policy, Is my data used for model training?, How long do you store my data?, model improvement settings
- OpenAI, ChatGPT pricing: plan prices, models by plan, context windows, feature table, training opt-out
- OpenAI, API models with the GPT-6 Astra and GPT-6 Sol pages: prices, context, cutoffs
- OpenAI, GPT-6 Astra and Introducing GPT-6 Sol and Luna: dates, availability, benchmark claims, figure
- OpenAI Help Center, GPT-6 and other models in ChatGPT: which model each plan runs, GPT-6 Pro allowances
- OpenAI Help Center, File Uploads FAQ, Projects in ChatGPT, Memory in ChatGPT, ChatGPT Voice
- OpenAI Help Center, How your data is used to improve model performance and Data controls in ChatGPT
- OpenAI Help Center, ChatGPT release notes (September and October 2026 entries: GPT-6 Astra, Images 2.5, ChatGPT for Word, Voice limits, GPT-6 in Chat, Pro 200 allowance) and OpenAI, Download ChatGPT
- OpenAI, Introducing GPT-6.1 Sol: availability in Work and Codex, API price, benchmark claims
- OpenAI Help Center, About ChatGPT Pro tiers: Pro 100, Pro 200 and Pro 500 prices, Pro 200 reopening and allowance, Ultrafast
- OpenAI, DevDay 2026 Recap: Pro 500 allowance, 29 September announcements
- OpenAI, Introducing dots: what dots are, who gets them, excluded regions
- Anthropic, Introducing Claude Sonnet 5.5: release date, API price, benchmark claims
