What Is a Prompt?

Updated November 2, 2026 — Le Chat is now Vibe. As of May 28, 2026, Le Chat is now Vibe (still at chat.mistral.ai): a single agent for work and code, with your conversations, settings, and subscriptions carried over. The legacy Le Chat experience remains accessible in Chat mode; in Work mode, agents are now skills (invoked with @), memories are the knowledge base, and deep research is the Deep Research skill. Plans are now called Mistral Free, Pro, Team, and Enterprise. This lesson now uses the name Vibe; some menus described here may have changed names or locations.

Updated November 2, 2026 — models. Mistral Large 3 (mistral-large-2512) and Mistral Medium 3.5 have been deprecated since October 31, 2026 and will be removed on April 30, 2027; their designated successor, Mistral Large 4 (mistral-large-4), has been in public preview since October 6. Examples in this course use the alias mistral-large-latest, which still points to Mistral Large 3 according to its official documentation (verified November 3, 2026), just as mistral-medium-latest points to Mistral Medium 3.5: these aliases will remain valid until April 30, 2027. For a new project, test mistral-large-4, their designated successor.

The prompt: your interface with artificial intelligence

When you interact with a Mistral language model, every message you send constitutes a prompt. This term, borrowed from English, refers to the instruction or request you submit to the model to get a response. Understanding what a prompt is and how it works is the first step to getting the most out of Mistral models, whether you use Vibe or the API.

A prompt can be as simple as a direct question—“What is the capital of France?”—or as elaborate as a multi-paragraph document containing context, examples, format constraints, and detailed instructions. Between these two extremes lies a whole spectrum, and you choose where to place the slider based on the stakes. The quality of the response you get depends directly on the quality of the prompt you formulate: this is the central thesis of this entire course, and you will verify it lesson after lesson.

How the model processes your prompt

To prompt well, you first need to understand what happens on the model side. Mistral models work by text completion: the model receives your prompt and generates the most probable continuation, token by token. A token generally corresponds to a word or a word fragment—“hello” might be a single token, while a rare word will be split into several pieces.

Concretely, the cycle is always the same: you send your prompt via Mistral’s Chat Completions API, the model analyzes all the provided context, then it generates its response token by token, choosing the most appropriate token at each step given everything that precedes it. The complete response is then returned to you in one block, or via streaming as it generates.

This mechanism explains why your prompt formulation is so important. The model does not “understand” your hidden intent: it extends the text you give it. If that text is vague, the most probable continuation will be equally vague text. If the text sets a precise framework, the most probable continuation will respect that framework. Your prompt literally determines the direction the generation will take.

The impact of the prompt on response quality

The difference between a good and a bad prompt can be spectacular, and nothing beats a direct comparison to convince yourself. Imagine you ask the model:

Tell me about marketing.

You will get a generic essay on what marketing is—accurate, but useless for your day-to-day work. Now compare that with this formulation:

You are a digital marketing consultant specializing in French SMEs.
List 5 content marketing strategies adapted to a budget of less than
€500 per month. For each strategy, indicate:
- The main channel
- The weekly time required
- The expected ROI at 6 months

Here, the model knows who it is supposed to be, who it is addressing, what budget constraint it is working under, and what format it must deliver. It will generate a concrete plan, structured and directly usable in a meeting. Keep this in mind: the difference between these two results does not lie with the model, which is the same in both cases, but solely with the precision of the instruction.

The components of an effective prompt

Breaking down the second prompt above, you find the building blocks you will manipulate throughout this course. Context provides the background information necessary to understand the task—in our example, the fact that we are talking about French SMEs with a small budget. Instruction states precisely what you expect from the model: list five strategies, not vague advice or a lecture. Format describes the desired structure for the response—list, table, JSON, or paragraph—and this is what makes the result usable without editing. Constraints set the limits to respect, such as length, tone, or vocabulary. Finally, examples show the model a demonstration of the expected result, a technique so effective that an entire lesson is dedicated to it later.

You do not need to include all these elements every time. A simple question requires only a clear instruction, and overloading a trivial prompt would be counter-productive. However, as soon as the task becomes complex—writing for a specific audience, analyzing data, producing a formatted deliverable—combining these components improves the result considerably.

Mistral’s Chat Completions API

With Mistral, prompts go through the Chat Completions API. This API uses a structured message format where each message has a role (system, user, assistant) and content. We will dive deeper into these roles in the next lesson; for now, just see what a minimal call looks like with the Mistral Python SDK:

from mistralai.client import Mistral

client = Mistral(api_key="your-api-key")

response = client.chat.complete(
    model="mistral-large-latest",
    messages=[
        {"role": "user", "content": "What are the 3 principles of effective prompting?"}
    ]
)

print(response.choices[0].message.content)

This code sends a simple prompt and displays the model’s response. Nothing more—and yet everything is already there: a message, a role, content. In the following lessons, you will learn to fully exploit the richness of this API to turn these three lines into real applications.

Key takeaways

• A prompt is the instruction you send to the Mistral model to get a response
• Prompt quality directly determines response quality
• An effective prompt combines context, instruction, format, and constraints
• Mistral models process prompts via the Chat Completions API
• Investing time in formulating your prompts is the most powerful lever to improve your results