Osum · Updated
Claude vs ChatGPT vs Gemini (2026): Which Wins?
Claude, ChatGPT and Gemini all write well enough to fool a busy reader. That is exactly why the choice is hard: the first draft always looks fine. The differences that decide your week are quieter. What each assistant keeps between chats, how its usage runs out, and what you can check before the draft leaves under your name. Here is how they compare on drafting, context and cost, with plan prices checked on the official pages, plus where each one breaks.
The short answer
Pick Claude if your week is real writing for clients or your team: updates, replies, proposals, summaries of your own documents. Its Projects hold your standing context, its plan pages list memory across conversations, and the writing keeps a tone professionals trust. Pick ChatGPT if you want the widest tool bench, from image generation to its coding and work modes, for one subscription. Pick Gemini if your work already lives in Google Workspace and you want the assistant that sits next to it. All three run a capable free tier.
The week-two truth is the same for all of them: the demo draft is easy, the standing setup is the job. An assistant saves you time only when your context, your voice and your checks survive the chat, and something runs over the draft before it goes out.
| Job | What users report | Your move |
|---|---|---|
| A recurring weekly document | Drafts that drift when the chat starts from zero: “I’ve never asked Claude to re explain things so many times with this.” | Put the context in a Project and brief it with a done-when line. |
| A claim that must be true | Confident lines with nothing behind them: “It would confidently declare things which any C++ programmer could tell you were bollocks straight away.” | Require a source beside every number, name and quote. |
| A heavy analysis ask | Usage that vanishes on the wrong setting: Anthropic’s own guidance ties the burn to model choice and effort. | Give the heavy ask a fresh thinking chat; run formatting on the light setting. |
| A draft under your name | Readers who clock the voice: “That is not your voice. That is the most Opus 5 writing imaginable.” | Keep a style file from your own writing and check the draft against it. |
What’s new in Claude for 2026
Anthropic’s Claude is no longer just a chat window. As of 28 September 2026, claude.com’s plan pages list memory across conversations, projects for organizing chats and documents, extended thinking, web search, file and code work, connectors with remote MCP, and desktop extensions as part of the product, and the Pro plan bundles Claude Code, Claude Cowork, Claude Design and Claude Science alongside more usage.
The usage story also changed in 2026. In April, after weeks of complaints, Anthropic’s product lead explained why sessions were draining limits faster than expected: tighter limits at peak hours, costlier long-context sessions, and a model effect, with Opus burning limits roughly twice as fast as Sonnet at the start of a session. The same guidance told users to start fresh sessions for chats idle about an hour and to lower effort or turn off extended thinking on light tasks. If your comparison of Claude predates April 2026, its usage advice is out of date.
Anthropic’s model line now spans several families, listed on claude.com as Mythos, Fable, Opus, Sonnet and Haiku. Release news lands on the company blog with dates, and that is the page to watch before any plan or price in this article goes stale.
Which Claude? The app, Claude Code and Claude Cowork
Claude today is Anthropic’s assistant: the chat app on the web, desktop and mobile. It is not Claude Code, the coding agent that works in your repository, and it is not Claude Cowork, the mode for hands-off desk work on your files. Both of those ship alongside the app, and both are included in the Pro plan as of 28 September 2026.
The difference matters when you search for help. Complaints about sub-agents, compaction or tokens-per-hour usually describe Claude Code, not the app. If the work you want off your plate is writing and analysis, the app is the product, and the setup it rewards, Projects, a style file and a source check, is what this comparison and the course behind it teach.
Claude vs ChatGPT, Gemini and Copilot
The practical difference is what each assistant holds onto, and where your work already lives. Claude keeps standing context in Projects and, on Anthropic’s plan pages, memory across conversations. ChatGPT carries the widest feature bench and its own work modes. Gemini sits inside Google’s apps, where your docs and mail may already live, and Copilot sits inside Microsoft 365 the same way.
Whichever you pick, the drift problem travels with you. Users describe the same pattern on every assistant: strong first drafts, then context you keep re-explaining and details nobody checked. One Claude user ended up splitting the work across tools because of limits: “I end up using ChatGPT for general coding tasks because of the limited session/weekly limit Claude pro offers.” The setup you build matters more than the logo you pick.
| Option | Where it runs | Paid from | Best at | Watch out for |
|---|---|---|---|---|
| Claude | Web, desktop and mobile apps | Pro, $17 a month billed annually ($200 up front) or $20 monthly (checked September 28, 2026) | Client-ready drafting, long documents, standing context in Projects | Usage limits land by plan and model; heavy settings burn faster |
| ChatGPT | Web, desktop and mobile apps | Plus, $20 a month (Go $8; checked September 28, 2026) | The widest feature bench: images, files, work and coding modes in one subscription | Usage is shared across its modes; the bench can bury the writing job |
| Gemini | The Gemini app and Google Workspace | Free app; paid Google AI Pro and Ultra tiers (prices at sign-in; checked September 28, 2026) | Work that already lives in Google’s apps, plus Workspace-aware drafting | Plan prices and limits render at sign-in; check them before you compare |
| Microsoft Copilot | Microsoft 365 apps | Included with many Microsoft 365 business plans (checked on Microsoft’s pages, September 28, 2026) | Drafting inside Word, Outlook and Teams for organizations already on Microsoft | Behaviour varies by tenant and plan; admin settings decide what it can touch |
Claude vs ChatGPT: who keeps the work
ChatGPT brings the bigger bench. Image generation, file analysis, agent-style work modes and a large ecosystem all ride on one subscription, and its plans run from Free through Go ($8), Plus ($20) and Pro (from $100) as of 28 September 2026. For tinkerers and teams who want one tool that attempts everything, that breadth is real value.
Claude is the narrower product aimed squarely at professional writing and analysis. Its Projects hold the standing context a weekly document needs, its plan pages list memory across conversations, and its April 2026 usage guidance is public, which makes the burn something you can plan around: heavy asks on the strong model in a fresh session, formatting on the light one. Users who left for ChatGPT often left because of limits, not output quality: “I end up using ChatGPT for general coding tasks because of the limited session/weekly limit Claude pro offers.”
Choose ChatGPT if you want one subscription that attempts every kind of job and you will manage the context yourself. Choose Claude if the job is a recurring document that must sound like you, with sources you can check, and you want the setup, Projects, style file, review, that keeps it standing. Whichever you pick, a brief with a done-when line decides the draft.
Claude vs Gemini: where your work already lives
Gemini’s advantage is the address it lives at. If your documents, mail and calendar are in Google Workspace, the Gemini app and its Workspace integrations sit next to the work already, and the app itself is free with paid Google AI Pro and Ultra tiers above it. Google publishes those plans on its AI plans page, with exact prices shown at sign-in, so check them there before comparing costs.
Claude’s advantage is the standing setup around the draft. Projects keep the Update’s sample, brief and notes in one place, the plan pages list memory and connectors, and Anthropic’s public usage guidance tells you where the allowance actually goes. For a document you rebuild every week under your own name, that setup, not the address, decides how much rewriting survives.
Choose Gemini if your week already runs through Google’s apps and you want the assistant beside them, starting free. Choose Claude if the deliverable is checked client-ready writing and you will invest an hour once in a setup that carries the context forward. Whichever you pick, the check before sending is yours.
What Claude costs right now
Claude has five plans: Free, Pro, Max, Team and Enterprise. These prices were checked on claude.com on September 28, 2026, and Anthropic’s page notes that prices and plans are subject to change.
Usage limits apply on every plan and vary with the model and the settings you run. Anthropic’s own April 2026 guidance ties the burn to model choice, with Opus draining limits roughly twice as fast as Sonnet at the start of a session, to effort settings and extended thinking, and to resuming sessions that have been idle for about an hour. If a limit keeps landing mid-week, the sort in that guidance is the first fix, before any upgrade.
| Plan | Price | What it includes |
|---|---|---|
| Free | $0 | Chat on web, desktop and mobile, with usage limits; the plan pages list memory, projects, extended thinking, web search and file work |
| Pro | $17 a month billed annually ($200 up front), or $20 monthly | More usage, unlimited projects, Research, memory, and Claude Code, Claude Cowork, Claude Design and Claude Science included |
| Max | From $100 a month | 5-20x more usage than Pro, higher output limits, early access to advanced features, priority at peak times |
| Team and Enterprise | From Anthropic’s sales pages | Collaboration, admin controls and security for organizations |
Where Claude breaks after the first days
The first break is context. The demo chat went well because you pasted everything into it. By the third chat you are re-explaining the project, the voice and last week’s decisions, and one user summed the cost up: “I’ve never asked Claude to re explain things so many times with this.” The fix is not a longer prompt. It is a standing Project with your context and a style file in it, briefed with a done-when line.
The second is the limit that lands mid-task. Anthropic’s April 2026 guidance confirmed limits tighten at peak hours and burn varies by model and session age: Opus drains roughly twice as fast as Sonnet at the start of a session, and a chat idle for about an hour costs more to resume than a fresh one. “I did huge amounts of work without hitting the limit - and small tasks at some other times that hit the limit before it completed,” one user wrote. The fix is the sort: analysis on the strong model in a fresh chat, formatting on the light setting.
The third is the confident wrong detail. “It would confidently declare things which any C++ programmer could tell you were bollocks straight away,” one long-time user wrote, and the same failure wears a suit in business writing: a forecast called a measurement, an audit that closed on a call still unscheduled. The fix is a source check before anything leaves, and a review that runs in order.
- Context: put the standing material in a Project and brief every ask with a done-when line.
- Limits: sort every ask by effort; give the heavy one a fresh, thinking chat.
- Accuracy: require a source beside every number, name and quote, and open it yourself.
The draft-in-your-voice template
One of the failure patterns above ends on a fix you can run today, and here it is: the lesson 1 brief from the Claude course. It takes one sample of your own writing and about 14 minutes, and it ends with a draft you have checked line by line.
The brief, word for word: “Here is how I sound: [paste a short sample of your own writing on the same kind of job]. Draft [the piece: this week’s update opener, a reply, a short section] in the same voice. Two sentences. No new facts.”
When the draft comes back, read it beside the sample: same length, same tone, same habits. Cut anything Claude added that you would never write. Nothing was sent anywhere; the draft waits in your notes until you decide.
No sample handy? Write the three habits you would defend in front of a client: your sentence length, the words you never use, how you end a piece. That list is the start of the style file lesson 5 builds.
- The brief: sample, piece, voice, two sentences, no new facts.
- The check: read beside the sample and cut what you would never write.
- The limit: this template drafts; it does not check facts. That check is lesson 3.
Claude FAQ
Is Claude worth paying for? If a recurring document eats your week, the Pro plan’s higher allowance and included tools pay for themselves fast; Pro was $17 a month billed annually or $20 monthly when we checked on 28 September 2026. The Free plan still runs real work, just under tighter usage limits.
How much does Claude cost? Five plans: Free at $0, Pro at $17 a month billed annually ($200 up front) or $20 monthly, Max from $100 a month for 5-20x Pro usage, plus Team and Enterprise tiers. Checked on claude.com on September 28, 2026.
Why does Claude keep forgetting my project? Chats start from zero by design. The fix is a Project: your sample, brief and notes stay attached to the work, so every chat inside it starts from your context.
Why did my Claude limit run out so fast? Burn varies with the model, effort settings and session age. Anthropic’s April 2026 guidance says Opus drains limits about twice as fast as Sonnet at session start and recommends fresh sessions for chats idle about an hour.
Can Claude write in my voice? Closer than any description can, if you show it a sample of your own writing and keep the rules in a style file. It still adds habits you did not write, so the read beside the sample is the check.
Is Claude better than ChatGPT? For a recurring professional document with standing context, many writers say yes. For the widest single-subscription feature bench, ChatGPT. Both run a capable free tier.
Is the Claude course made by Anthropic? No. Osum is independent and not affiliated with Anthropic, and the certificate is an Osum certificate.
Verdict: which assistant should you hand your writing to?
Claude is the pick for professionals whose week is a real document: the consultant’s Monday update, the analyst’s summary, the freelancer’s client replies. Projects hold the standing context, the plan pages list memory across the chats, and the usage guidance is public enough to plan around. Choose ChatGPT if you want the widest bench on one subscription. Choose Gemini if your work already lives in Google’s apps and you want the assistant beside them, starting free.
Whichever you pick, the skill that decides whether it saves you time is the same: a standing setup with your context and your voice in it, a brief that says what done means, and a check on every claim before the draft leaves. If you’re going with Claude, our course teaches exactly that on your own writing in 12 hands-on lessons, with a certificate and every other Osum AI course included.