---
title: Chat Completion with Ministral 7B
slug: chat-completion-with-ministral-7b
url: https://apyhub.com/mastera-ai/service/chat-completion-with-ministral-7b
provider: MasterA AI
auth: api_key
version: v1
service_type: sync
endpoints: 1
atoms: 1
mcp: true
alias: chat-completion-with-ministral-8b
---

# Chat Completion with Ministral 7B

Chat completions API powered by Ministral 7B on a dedicated GPU. OpenAI-compatible request/response format, system role support, fast synchronous replies.

## Endpoints

| Method | URL | Description | Atoms |
| --- | --- | --- | --- |
| POST | `https://api.eu.apyhub.com/mastera-ai/chat-completion-with-ministral-7b/v1/mistral/chat/completions` | What it does OpenAI-compatible chat completions using Ministral 7B. Send model ("mistral-7b"), a me… | 1 |

## Endpoint reference

### Chat completion (Ministral 7B)

`POST https://api.eu.apyhub.com/mastera-ai/chat-completion-with-ministral-7b/v1/mistral/chat/completions` · 1 atoms · accepts `application/json` · returns `application/json`

| Parameter | In | Type | Required | Description |
| --- | --- | --- | --- | --- |
| `model` | body | string | yes | Must be mistral-7b. One of: mistral-7b. Example: `mistral-7b`. |
| `messages` | body | array of object | yes | OpenAI-style chat messages. Supports system, user, and assistant roles. |
| `messages[].role` | body | string | yes | One of: system, user, assistant. |
| `messages[].content` | body | string | yes |  |
| `max_tokens` | body | integer | no | Maximum tokens to generate. Prompt plus max_tokens must fit the 6,144-token context window. Default: `512`. Example: `200`. |
| `temperature` | body | number | no | Sampling temperature. Default: `0.7`. Example: `0.7`. |

#### Quickstart

```bash
curl -X POST "https://api.eu.apyhub.com/mastera-ai/chat-completion-with-ministral-7b/v1/mistral/chat/completions" \
  -H "apy-token: $APY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "mistral-7b",
    "messages": [
      {"role": "system", "content": "You are a helpful assistant."},
      {"role": "user", "content": "Write a short haiku about the ocean."}
    ],
    "max_tokens": 200,
    "temperature": 0.7
  }'
```

#### What you'll get back

```json
{
  "id": "chatcmpl-ae8e9f07da25157e",
  "object": "chat.completion",
  "created": 1790318354,
  "model": "mistral-7b",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Deep blue whisper\nSalt-kissed waves embrace\nEternal tide.",
        "refusal": null,
        "annotations": null,
        "audio": null,
        "function_call": null,
        "tool_calls": [],
        "reasoning": null
      },
      "logprobs": null,
      "finish_reason": "stop",
      "stop_reason": null,
      "token_ids": null,
      "routed_experts": null
    }
  ],
  "service_tier": null,
  "system_fingerprint": "vllm-0.23.0-bfb893d8",
  "usage": {
    "prompt_tokens": 18,
    "total_tokens": 34,
    "completion_tokens": 16,
    "prompt_tokens_details": null
  },
  "prompt_logprobs": null,
  "prompt_token_ids": null,
  "prompt_text": null,
  "kv_transfer_params": null
}
```

## About

## What it does
Ministral Chat Completion generates replies from a conversation using Ministral 7B Instruct, with the same request and response shape as the OpenAI Chat Completions API. Send a `messages` array of `system`, `user`, and `assistant` turns with `model` set to `mistral-7b`, and the generated reply comes back in a single synchronous call. There are no job IDs and no polling.

You can tune each call with `max_tokens` (default 512) and `temperature` (0 to 2, default 0.7). The reply is in `choices[].message.content`. The `finish_reason` field is `stop` or `length`, so you can tell when output was cut off. A `usage` object gives prompt, completion, and total token counts for tracking consumption.

Use it for chatbots and conversational agents, drafting and rewriting copy, summarising short text, Q&A over content you pass in, and other general language tasks where a fast reply matters more than frontier-model depth. If your code already builds OpenAI-style chat payloads, you can send them as they are.

> **Note:** Runs Ministral-7B-Instruct-2410 (AWQ quantized) on a dedicated NVIDIA RTX 4070 Ti SUPER via vLLM. Typical response time is 1–3 seconds, depending on `max_tokens`.

## Usage

Base URL: `https://api.eu.apyhub.com` (default region — see
`GET https://apyhub.com/api/public/regions` for the rest).

Authenticate with an ApyHub API key in the `apy-token` header.
Full docs and a live playground: https://apyhub.com/mastera-ai/service/chat-completion-with-ministral-7b