Claude vs. ChatGPT: the head-to-head comparison
A tenth of a point separates the top two in our ranking. Where Claude leads, where ChatGPT does, what the plans cost, and what changed in September with GPT-6 and Claude Opus 5.5.
Our test result in figures
In our overall ranking Claude scores 4.5 out of 5, just ahead of ChatGPT at 4.4. The gap is unevenly spread: Claude is slightly ahead on output quality (4.8 to 4.7) and ease of use (4.7 to 4.6), ChatGPT clearly ahead on features (4.9 to 4.4) and narrowly on value for money (4.3 to 4.2). Data protection tips the balance, 4.2 to 3.6.
A note first: our scores come from the test in August 2026. Since then OpenAI has shipped GPT-6, first for paying customers and, since 22 September, in a smaller version on the free plan too; Anthropic released Claude Opus 5.5 on the same day. We are re-testing both. The fundamental differences this comparison is about, however, depend less on the individual model than on focus, features and pricing.
Texts and long documents: Claude
With long documents the gap in our assistant comparison was at its widest. On a 40-page tender, Claude connected details in the appendix with the deadlines at the front and pointed out an inconsistency of its own accord; ChatGPT delivered a readable version but left out two secondary conditions.
The context window fits the picture: according to the vendor, current Claude models handle up to one million tokens in the chat, enough for several thick reports at once. And in terms of language, Claude delivers what we consider the cleanest German prose in the field: few stock phrases, clear paragraphs, a tone that rarely needs correcting.
Breadth of features: ChatGPT
ChatGPT is the broader tool. It creates images, works through spreadsheets with real code execution rather than estimating figures, holds voice conversations, researches in several steps and completes tasks in the browser as an agent. In our test its analysis of a revenue table was error-free and traceable.
Claude caught up in 2026: web search, in-depth research, memory across conversations, projects, file creation and, since the end of August, the Claude in Chrome browser control on all paid plans. One gap remains: Claude does not create photos or illustrations, only graphics such as diagrams. If you need images, use ChatGPT or a specialist tool from the AI image generators category.
Coding: Claude Code versus Codex
For coding, both vendors field their own tools. Claude Code works in the terminal and the editor directly inside the project and is included from the Pro plan; in our AI coding tools category it scores 4.3. OpenAI offers a comparable tool, Codex, on ChatGPT's paid plans.
For developers this is often the real deciding factor: if you subscribe for the coding tool anyway, you get the assistant for everything else thrown in.
Plans and limits
Claude is available free with a message allowance that renews every five hours; Claude Code is not included on the free plan. Pro costs 20 dollars a month or 17 dollars on annual billing, and the Max plans with five or twenty times the allowance cost 100 and 200 dollars. Anthropic bills in dollars, plus VAT.
In August 2026 ChatGPT lifted the limits on text conversations on its free plan, but since 24 August it has shown ads there in Germany, as it does on the low-cost Go plan at around 8 euros. Plus costs around 23 euros a month in Germany and stays ad-free; above it are two Pro tiers at 100 and 200 dollars. For teams, both vendors charge around 25 dollars per person per month.
For most individual users the sum is therefore the same: 20 dollars for Claude Pro or around 23 euros for ChatGPT Plus. Occasional writers get by with either free plan, but have to accept ads on ChatGPT.
Data protection: Claude ahead, but no longer automatically
At ChatGPT, using conversations for training is switched on by default on the Free, Go, Plus and Pro plans and has to be turned off in the settings. Anthropic changed its rules for consumers in autumn 2025: since then a switch decides whether conversations flow into training. If it is on, Anthropic keeps the data for up to five years, otherwise 30 days. Anyone who simply confirmed the notice at the time should check the setting.
On the team and enterprise plans, training use is excluded at both vendors, and only there is a data processing agreement available. Neither offers EU data storage to consumers. We see Claude's remaining lead in the clearer default and the shorter retention without training; what needs clarifying before business use is in our GDPR guide.
Which assistant for whom
Claude is the better choice for knowledge work with text: contracts, studies, reports, editing, and for anyone who also codes. ChatGPT is the better choice for anyone who wants one tool for as many tasks as possible, needs images, works a lot with figures or likes to ask by voice.
The most honest recommendation remains the one from our big comparison: use both. The free plans are enough to get a second opinion on important tasks. If you are considering entirely different assistants, you will find them in our article on ChatGPT alternatives.
In short: Claude (4.5) ahead of ChatGPT (4.4). Claude for long documents, German texts and coding, ChatGPT for breadth of features, images and figures. Both cost around 20 dollars a month; ChatGPT shows ads on its free plan. Check training use in the settings of both.
Tools discussed in this article
Each tool has a full review with scores and pricing.
Jean-Marc Mihoc works with AI tools every day and tests them here on real tasks, with accounts he pays for himself, rather than on demos. He writes down what comes out of it, even when that is unspectacular, and says for every tool who should not bother with it.
