The short answer
ChatGPT, from OpenAI, is the better all-round choice: it wins writing, coding, research, files and the free plan, with GPT-6 Sol on Plus and higher plans. Vibe, the Mistral AI assistant formerly called Le Chat (renamed on 28 May 2026), wins on data location, connected-tool agents and self-hosting: it hosts data in the EU by default, asks before acting in your tools, and Enterprise can run on-premises. EU teams with data protection concerns should start with Vibe.
| ChatGPT | Vibe (Le Chat) | |
|---|---|---|
| Made by | OpenAI, United States | Mistral AI, Paris, France |
| Free plan | Unlimited text chats with GPT-6 Luna, limited extras | Capped messages and searches, 100+ connectors, 5 scheduled tasks |
| Models | GPT-6 Luna, GPT-6 Sol; GPT-6.1 Sol and Terra in Work and Codex | Mistral models; Mistral does not say which powers each mode |
| Paid plans | Go, Plus, Pro, Business, Enterprise | Pro, Education, Team, Enterprise |
| Agent mode | ChatGPT Work | Vibe Work, asks for confirmation before sensitive actions |
| Coding | Codex | Vibe Code: sandboxes, CLI, VS Code, JetBrains, Zed |
| Your documents | Projects on paid plans | Libraries stored in Mistral's European cloud |
| Where data is hosted | OpenAI, a US company; EU data residency only for new Enterprise and Edu workspaces | EU by default |
| Training on your chats | On by default for Free, Go, Plus, Pro; off for Business and Enterprise | On by default for individuals; Enterprise opted out |
| Self-hosting | Not offered for ChatGPT | Enterprise on-premises; open-weight Mistral Medium 3.5 |
ChatGPT vs Vibe (Le Chat): the verdict
Pick ChatGPT if you want the strongest all-round assistant: it wins writing, coding, research, files and the free plan, and OpenAI tells you exactly which model you get on each plan. Pick Vibe, the assistant Mistral AI used to call Le Chat, if where your data lives matters: it is hosted in the EU by default, it asks before it sends, writes or changes anything in your connected tools, and Mistral publishes open-weight models your company can run itself. For a European team with data protection worries, Vibe is the sensible default; for everyone else, ChatGPT is.
A word on the name. Mistral announced on 28 May 2026 that Vibe replaces Le Chat, combining everyday work and coding in one agent, and its help centre confirmed the change in August. The address chat.mistral.ai still works and opens Vibe with the same login and conversations. Vibe has three modes: Work for multi-step tasks across your tools, Code for coding sessions, and Chat for turn-based conversation. Many guides online still describe the old Le Chat, so menu names you read elsewhere may be out of date.
ChatGPT, for comparison, runs GPT-6 Luna for Free and Go users in the Chat tab and GPT-6 Sol for Plus, Pro, Business and Enterprise since 7 October 2026, with GPT-6.1 Sol and GPT-6.1 Terra in ChatGPT Work and Codex on Plus and higher plans.
Writing: ChatGPT wins
ChatGPT wins because you know what you are getting and the tools around the draft are deeper. Paid plans write with GPT-6 Sol, keep context in projects and memory, and can turn a draft into a deck with ChatGPT Slides (eligible Pro, Business and Enterprise users). Vibe writes well and has Canvas for editing a document alongside the chat, but Mistral does not say which model powers each Vibe mode, which makes it harder to know what you are comparing. If you write in French or other European languages, try both on your own text before you decide.
Coding: ChatGPT wins on power, Vibe on what Pro includes
ChatGPT's Codex runs OpenAI's newest models, GPT-6.1 Sol and GPT-6.1 Terra, on Plus and higher plans, and Business connects to GitHub. That is the stronger setup for a team that codes every day.
Vibe is closer than its reputation suggests. Vibe Code runs remote coding sessions in isolated sandboxes that can end in a pull request, and it works in a command-line tool (powered by Mistral's Devstral model), VS Code, JetBrains and Zed. Mistral's Pro plan includes all-day coding in the CLI, IDE or web in the same subscription as everyday work, and the Free plan includes limited coding sessions. Our pick is still ChatGPT for coding, because Codex runs OpenAI's newest GPT-6.1 models, but Vibe is a fair choice for a European developer who wants one subscription.
Research: ChatGPT wins for the web, Vibe for your own documents
For web research, ChatGPT wins: deep research is on every plan with higher limits on paid plans. Vibe offers deep research reports on Pro, and its Libraries feature is a real strength: you build indexed collections of your own documents and web pages, Work cites from them, and they are stored in Mistral's European cloud. If your research is mostly your own reports and policies, Vibe's Libraries are a good fit. For cited web research, also look at Perplexity.
Files and data: ChatGPT wins
ChatGPT analyses spreadsheets, creates documents, decks and spreadsheets in ChatGPT Work, has extensions for Excel, Word, PowerPoint and Google Sheets, and transcribes uploaded audio on paid plans. Vibe Work lists files, Canvas, spreadsheets and Code Interpreter, but Code Interpreter, which runs the actual calculations, is on paid plans only, and spreadsheet and slide generation is called out on the Education plan. ChatGPT's file tools are broader and better documented.
Connected tools and agents: Vibe wins
Vibe connects to more than 100 tools, including email, calendar, Slack, GitHub, Jira, Notion, Google Drive and SharePoint, and the Free plan lists those connectors. Vibe Work breaks the job into steps, shows its tool calls and progress, and asks for confirmation before sensitive actions such as sending email, posting messages or changing data in external tools. That approval step is exactly what a cautious team wants from an agent. ChatGPT has iCloud Mail and Calendar plugins on all plans and connects Google Workspace, Slack, GitHub and Microsoft 365 on Business, and ChatGPT Work also acts in connected apps. For free access to many connectors with a human check built in, Vibe wins.
Privacy and data location: Vibe wins
- Vibe. Mistral AI is based in Paris and hosts data in the EU by default, with some temporary transfers through subprocessors it lists. Individual users' input and output are used for training by default until you opt out: on mobile go to Settings > Account > Data & Account Controls and untick Enable data sharing. On Team, admins can switch training off for the whole organisation in Manage > Vibe > Privacy. Enterprise is opted out by default and can be deployed privately or on-premises.
- ChatGPT. OpenAI is a US company. Training is on by default for Free, Go, Plus and Pro; switch it off in Settings > Data controls > Improve the model for everyone, or use a temporary chat. Business, Enterprise and Edu content is not used for training by default. New ChatGPT Enterprise and Edu workspaces can choose to store content at rest in Europe.
Both train on personal plans until you opt out, so switch the setting off whichever you choose. Vibe wins the category on data location: EU hosting by default plus an on-premises option is something ChatGPT does not offer; OpenAI's EU data residency is limited to new Enterprise and Edu workspaces.
Price and value: ChatGPT wins on the free plan
ChatGPT Free gives unlimited text chats with GPT-6 Luna, within abuse guardrails, plus limited uploads, images, voice, deep research and Codex. Vibe Free caps messages and web searches but adds 100+ connectors, up to five scheduled tasks and limited coding sessions. For most people who mainly chat, ChatGPT's free plan goes further. On paid plans, compare the current prices on ChatGPT's pricing page and Mistral's pricing page; Mistral's Education plan gives verified students Pro for up to 12 months.
Open weights and self-hosting: Vibe wins
Mistral launched Mistral Medium 3.5, a 128-billion-parameter open-weight model with a 256k context, alongside Work mode in May 2026, and Vibe Enterprise can run privately or on-premises. OpenAI's open-weight models, gpt-oss-120b and gpt-oss-20b, were last updated on Hugging Face in August 2025 and are not the GPT-6 models in ChatGPT. If you want the assistant and the model inside your own walls, Mistral is the only one of the two that offers it as a product.
Same task, both tools
These comparisons describe each tool's documented features, not our own test runs.
Prepare for a client meeting from your inbox and drive
Vibe: in Work mode, ask it to gather the latest emails and files about the client from your connected email and Google Drive and write a one-page brief; it shows its tool calls and asks before changing anything. ChatGPT: on Business, connect Google Workspace or Microsoft 365 and ask ChatGPT Work for the same brief. On a personal plan, iCloud Mail is the documented connection. Edge: Vibe for personal and small-team users, level on business plans.
Analyse a spreadsheet of monthly costs
ChatGPT: upload the file and ask for the three biggest changes with a chart; data analysis runs the numbers. Vibe: the same request on a paid plan uses Code Interpreter; on Free, there is no Code Interpreter. Edge: ChatGPT.
Answer staff questions from the HR handbook
Vibe: add the handbook to a Library, which is stored in Mistral's European cloud, and Work cites the passages it uses. ChatGPT: put the handbook in a project on a paid plan and ask questions there. Edge: Vibe for EU teams.
Who should pick ChatGPT, and who should pick Vibe
Pick ChatGPT if you want the most capable general assistant, heavy file and spreadsheet work, the newest OpenAI models for coding, or the most generous free chat. Read how to use ChatGPT.
Pick Vibe if you work in the EU or for a client who cares about EU hosting, you want an agent that asks before it acts in your email and documents, or your company may want to self-host later. Read how to use Vibe (formerly Le Chat).
Who should skip Vibe: anyone who needs to know exactly which model answered, since Mistral does not say which model powers each mode. Who should skip ChatGPT: teams that need on-premises deployment. For more options, see the best ChatGPT alternatives, ChatGPT vs DeepSeek or our comparison tool, and practise in the free lessons at /learn?tool=lechat.
Prompts to copy
Build a client brief from your tools
Find the last [number] emails and any documents about [client or project] in my connected [email and drive]. Write a one-page brief with the current status, open questions, risks and the three things I should raise in our next meeting. List every source you used. Do not send or change anything.
You get: A sourced meeting brief, with the agent told not to act (Vibe Work or ChatGPT Work)
Answer questions from your own documents
Using only the documents in [library or project name], answer this question: [question]. Quote the passage you rely on and name the document. If the documents do not answer it, say so instead of guessing.
You get: An answer grounded in your own files, with the passage quoted
Find the story in a spreadsheet
Here is a spreadsheet of [data]: [upload]. Find the three biggest changes between [period A] and [period B], show them in one chart, and write three sentences a manager could read in 30 seconds. Say which numbers you calculated and which you assumed.
You get: A chart and a short readout (needs data analysis in ChatGPT or Code Interpreter on a paid Vibe plan)