Both cost $20 a month and both answer almost anything you type. The difference is what runs out. ChatGPT never caps plain text chat and adds images and an $8 plan. Claude gives chat and its coding agent one shared allowance, makes no images, and shows no ads.
also searched as: claude vs chatgpt · chatgpt vs claude · claude ai vs chatgpt · claude or chatgpt · difference between claude and chatgpt · claude vs gpt · chatgpt or claude
Pick Claude if most of your use is long documents or code, you want the coding agent inside the same $20, and you would rather hit a limit than see an ad.
Pick ChatGPT if you want one app for everything, including pictures and voice, and you chat in volume, because plain text chats are not capped on any ChatGPT plan.
Pick neither if you would open it a few times a week, because both free plans already answer everyday questions and remember earlier chats, and $240 a year only buys headroom you would never touch.
prices checked 2026-10-05 · by dan scutaru
| Claude | ChatGPT | |
|---|---|---|
| price | free tier · from $20/mo | free tier · from $8/mo |
| free tier | Chat, web search, file creation, code execution, memory and up to 5 projects, on Sonnet and Haiku, with a rolling five-hour usage window. No Claude Code, no Research, no Opus or Fable models, no image generation. | Unlimited text chats with GPT-6 Luna (subject to abuse guardrails), with limited uploads, image generation, voice, deep research, memory and Codex access. |
| runs on | web, ios, android, mac, windows, linux | web, ios, android, mac, windows |
| the $20 plan | Pro. $20 billed monthly, or $200 upfront for a year ($17/mo). | Plus. $20 billed monthly. No annual billing on any personal plan. |
| plans above it | Max at $100 (5x Pro usage) or $200 (20x). | Pro at $100, $200 or $500. |
| usage limits | One allowance for chat and Claude Code. Five-hour session window plus a weekly limit on paid plans. No message counts published. | Everyday text chats uncapped on every plan, within abuse guardrails. Reasoning models, images, deep research and Codex are metered. |
| newest model on $20 | Fable 5 and 5.1 are outside Pro's limits. Pay-as-you-go credits only. | GPT-6 reasoning models are in Plus, with limits. GPT-6 Astra Pro reasoning is listed under Pro. |
Same price, different shape. Claude Pro adds Claude Code, Research, more models and unlimited projects to the free plan, which stops at five. ChatGPT Plus adds GPT-6 reasoning models, deep research, projects, scheduled tasks, custom GPTs and more Codex.
Two details decide it. Claude's newest models, Fable 5 and 5.1, are not in Pro's allowance. You pay for them with usage credits, and they are only bundled from Max up, capped at half the weekly limit. And only Claude sells a year upfront, which is cheaper until the plan changes under you.
Both bundle a coding agent with the subscription, so the real question is how long you can work before it stops. Claude Code needs Pro and spends the same allowance as your chats, on a five-hour window plus a weekly limit, with no published counts.
Codex comes with every ChatGPT plan, Free and Go included. On Plus, OpenAI estimates 15 to 150 messages per five hours on GPT-6 Sol and 5 to 45 on GPT-6 Astra, and says Pro currently has no five-hour limit. We print no benchmark. Give both the same real bug before you commit.
Most reviewers who compare the two prefer Claude's prose, and that is a taste call nobody can settle for you. What you can check is the workflow. Claude keeps a long draft in view and has no image tool to illustrate it. ChatGPT Plus reads roughly 40 pages in instant mode, about 320 with a reasoning model, and can make the header image too.
Paste three paragraphs of your own writing into both free plans and ask for an edit. Keep the one that changes less.
The model is rarely the complaint. The limits are. In a September 2026 Hacker News thread on default models, a former ChatGPT subscriber wrote that they wanted to like Claude but were "put off by reports of its tight usage limits". The same comment called ChatGPT's newest free model painfully rambling.
Notice what is missing. Almost nobody switches because one is smarter. They switch because they ran out, or got tired of the voice.
For long documents and coding inside one $20 plan, with no ads, yes. For images, uncapped text chat and a cheaper paid tier, no. The gap between the models is smaller than the gap between their limits.
Yes. The free plan covers chat on web, desktop and mobile, web search, file creation, code execution and memory. It leaves out Claude Code, Research and the Fable models, and usage resets on a five-hour window. The full answer →
No. Anthropic's help centre says Claude does not generate photos or illustrations. It reads images you upload and draws charts and diagrams in HTML and SVG. ChatGPT generates images on every plan, with a small allowance on Free.
Anthropic publishes no message count. Usage resets on a rolling five-hour session, paid plans add a weekly limit, and chat, the desktop app and Claude Code all spend the same allowance. Max multiplies Pro's session allowance by 5 or 20.