Claude vs ChatGPT: How They Differ for Work

Claude and ChatGPT are more alike than their fans suggest. Both are general AI assistants with a chat, projects that hold files and instructions, memory across conversations, connections to your other apps, an agent that takes on longer jobs and returns finished files, and a separate coding agent. Both have a Free plan. The differences are in how each piece works: ChatGPT splits its product into Chat, Work and Codex, while Claude is folding chat and Cowork into one conversation and keeps Claude Code alongside; a shared ChatGPT project includes its chats, while a shared Claude project, on Team and Enterprise, shares knowledge and instructions but keeps chats private; ChatGPT bundles app connections into plugins, while Claude calls them connectors. Neither wins everywhere, so the useful answer is which one fits which job. Everything below follows each vendor’s own documentation as of September 30, 2026.

Two narrower questions have their own pages: coding in the chat apps is in Claude vs ChatGPT for coding, and moving your setup from one to the other is in moving from ChatGPT to Claude. This page is the general comparison for work.

Claude vs ChatGPT at a glance

• Maker: Claude is made by Anthropic (what is Anthropic); ChatGPT by OpenAI.
• Where you use them: both on the web, in desktop apps and in mobile apps.
• Modes: OpenAI’s guide to using ChatGPT↗ (opens in a new tab) describes three, Chat for back-and-forth, ChatGPT Work for “a reviewable result”, and Codex for developer tools. Claude has chat, Cowork for longer tasks, and Claude Code for code.
• Plans, by name: Claude has Free, Pro, Max, Team and Enterprise. ChatGPT has Free, Go, Plus, Pro, Business, Enterprise and Edu.
• Model families: Claude runs on Fable, Opus, Sonnet and Haiku models. ChatGPT runs on OpenAI’s GPT models, with Astra, Sol and Luna as the current names.

Projects

Both use projects to keep the files, instructions and chats for one piece of work together, so you stop re-explaining it. Anthropic’s page What are projects?↗ (opens in a new tab) calls a Claude project a self-contained workspace with its own chat history and knowledge base, available on every plan, with Free accounts limited to five and paid plans able to search larger knowledge.

OpenAI’s page on projects and chats↗ (opens in a new tab) gives each ChatGPT project a Chats section and a Sources section for uploaded files and connected context, and says one project can hold chats started in Chat and in Work. On the web, a ChatGPT project has no direct access to a folder on your computer; the desktop app adds local projects that do.

The biggest difference is sharing. Claude shares projects on Team and Enterprise only, and colleagues see the knowledge and instructions; your chats in the project stay private unless you share them one by one. In a shared ChatGPT project, OpenAI Academy’s projects guide says members contribute directly in the project’s chats as well as its files and instructions, so the conversation itself is the shared record. Which ChatGPT plans can share, and how project-only memory works, are in Claude Projects vs ChatGPT Projects.

Memory

• Claude: memory is on by default on Free, Pro and Max and off by default on Team and Enterprise until an owner turns it on. Each project has its own memory, kept apart from other projects, and incognito chats are not remembered at all.
• ChatGPT: memory is managed under Settings, Personalization, and carries context from earlier chats into future work. OpenAI’s docs call memories a helpful recall layer, “not the only source for rules that must always apply.”

The practical rule is the same for both: anything that must be followed word for word belongs in instructions or project files, not in memory. If you are moving, Claude can import a list of memories you ask ChatGPT to write out, as moving from ChatGPT to Claude shows.

Connectors and plugins

Both reach your other tools through the Model Context Protocol, under different names. In Claude they are connectors: Anthropic’s connectors page↗ (opens in a new tab) describes a directory of ready-made ones and custom connectors to any remote MCP server, on every plan, with Free accounts limited to one custom connector.

In ChatGPT they arrive as plugins↗ (opens in a new tab), which OpenAI says “bundle capabilities into reusable workflows” and “can include skills and MCP servers”, working in Chat and Work on the web, desktop and mobile. To add your own MCP server, ChatGPT has developer mode, which OpenAI lists for Pro, Plus, Business, Enterprise and Education accounts on the web. What each side can reach is in Claude connectors explained and ChatGPT connectors and apps.

Files and finished output

Both read the files you upload and both make files you can download. Claude’s Free plan already includes file creation and code execution, and its artifacts turn a response into a document, dashboard or small tool beside the chat that you can share. ChatGPT Work is built around finished files, such as a deck, an analysis or a spreadsheet you review, and ChatGPT can preview and refine documents, presentations, spreadsheets and PDFs in the chat. How the two preview panes compare is in Claude artifacts vs ChatGPT Canvas.

Agents for longer jobs

Each has an agent that takes an outcome rather than a prompt, plans the steps, works in the cloud while you do something else, reaches your computer through its desktop app when a task needs local files, and can repeat work on a schedule. On the ChatGPT side that is ChatGPT Work, a mode you switch to. On the Claude side it is Cowork, on paid plans, and it is disappearing as a separate mode: Anthropic’s page Claude Cowork and chat are one Claude↗ (opens in a new tab) describes a merged experience rolling out from Pro and Max, where Claude decides whether a request needs an answer or a task. The detailed comparison, including approvals and triggers, is in ChatGPT Work vs Claude Cowork.

Coding, briefly

Both vendors send serious coding to a separate agent: Claude Code for Claude and Codex for ChatGPT. One detail matters when you choose on the chat alone: OpenAI’s models page says its newest models, GPT-6.1 Sol, GPT-6 Sol and GPT-6 Luna, are available in Work and Codex but not in Chat. The chat-app comparison is in Claude vs ChatGPT for coding, and the agents are compared in Claude Code vs Codex.

Claude or ChatGPT? Pick by job

• A group that wants to work in the same conversations: a shared ChatGPT project, where the chats are part of what is shared.
• A team on a business plan that wants shared reference material but private working chats: Claude Team or Enterprise.
• Reference material you will query all month, such as policies, a contract set or a product spec: either; both projects hold it, and Claude’s paid plans search large project knowledge.
• A task that ends in a file, such as a board-meeting deck or a vendor comparison spreadsheet: ChatGPT Work or Claude Cowork. Try the same task in both.
• Something small to show people, like a sales tax calculator for three states or a one-page dashboard: Claude artifacts, which you can share.
• Work that starts when something happens in Gmail, Slack or GitHub: ChatGPT, whose scheduled tasks can also run on events on eligible plans.
• Connecting a tool of your own on a Free plan: Claude, which allows one custom connector on Free; ChatGPT’s developer mode starts at Plus.
• Changes to a codebase: neither chat; use Claude Code or Codex.
• Moving from one to the other: one checklist, in moving from ChatGPT to Claude.

A 30-minute test on your own work

Reviews and benchmarks will not tell you which one suits your work. A short test will. Use the free plans, or the paid plans you are considering, and give both the same material.

1. Pick three real tasks from last week: one question, one rewrite and one job that ends in a file.
2. Create a project in each with the same two files and the same three lines of instructions.
3. Paste the same prompts, word for word, into both. Do not coach either one in follow-ups until both have answered.
4. Score each answer on what matters to you: accuracy against the files, whether it followed the instructions, and how much you had to fix.
5. Repeat the file task in ChatGPT Work and Claude Cowork if your plans include them.
6. Keep the one that needed less fixing for that kind of task. It can be different for different tasks.

Using the two attached files (our price list and last quarter's orders):
1. Summarize the five best-selling products by revenue.
2. Flag any order where the sales tax looks wrong for the ship-to state.
3. Draft a three-sentence note to the sales team about what changed.
Cite the file and row for every number. If something is missing, say so instead of guessing.

Using both on the same work

Many teams end up with both, because people use the assistant that already holds their context. That makes the list of work the thing to keep out of either one. fenbs is a small task board that both reach over MCP: Claude as a custom connector and ChatGPT as a developer-mode app where the plan allows it, each signing in through the browser and holding your role on the board. Each task is a feature, enhancement or bug with a note, a plan and a priority from 1 to 10, moving through To Do, Next Up, In Progress and Completed, and History records which assistant changed what, on whose behalf. fenbs has no due dates, sprints or assignee field, and it does not import from either assistant; it keeps one list both can read.