Usage

Scaffolding

Together with Devstral 2, we are releasing Mistral Vibe, a CLI tool allowing developers to leverage Devstral capabilities directly in your terminal.

• Mistral Vibe (recommended): Learn how to use it here

Devstral 2 can also be used with the following scaffoldings:

• Cline
• Kilo Code
• Claude Code
• OpenHands
• SWE Agent

You can use Devstral 2 either through our API or by running locally.

Mistral Vibe

The Mistral Vibe CLI is a command-line tool designed to help developers leverage Devstral’s capabilities directly from their terminal.

We recommend installing Mistral Vibe using uv for faster and more reliable dependency management:

uv tool install mistral-vibe

curl -LsSf https://mistral.ai/vibe/install.sh | sh

If you prefer using pip, use:

pip install mistral-vibe

To launch the CLI, navigate to your project's root directory and simply execute:

vibe

If this is your first time running Vibe, it will:

• Create a default configuration file at ~/.vibe/config.toml.
• Prompt you to enter your API key if it's not already configured, follow these instructions to create an Account and get an API key.
• Save your API key to ~/.vibe/.env for future use.

Local Deployment

The model can also be deployed with the following libraries:

• vllm (recommended): See here
• sglang: See here
• transformers: See here

We're thankful to the llama.cpp team and their community as well as the LM Studio and Ollama teams that worked hard to make these models also available for their frameworks.

You can now also run Devstral using these (alphabetical ordered) frameworks:

• llama.cpp: To use community ones such as Unsloth's or Bartowski's make sure to use changes from this PR.
• LM Studio: https://lmstudio.ai/models/devstral-2
• Ollama: https://ollama.com/library/devstral-small-2

If you notice subpar performance with local serving, please submit issues to the relevant framework so that it can be fixed and in the meantime we advise to use the Mistral AI API.

vLLM (recommended)

<details>
<summary>Expand</summary

We recommend using this model with the vLLM library
to implement production-ready inference pipelines.

Installation

Please make sure to install latest vLLM:

uv pip install -U vllm

Alternatively you can also directly use the latest docker image vllm/vllm-openai:latest:

docker pull vllm/vllm-openai:latest
docker run -it vllm/vllm-openai:latest

Also make sure to have installed mistral_common >= 1.8.6.
To check:

python -c "import mistral_common; print(mistral_common.__version__)"

Launch server

We recommand that you use Devstral in a server/client setting.

1. Spin up a server:

vllm serve mistralai/Devstral-Small-2-24B-Instruct-2512 \
    --max-model-len 262144 --tensor-parallel-size 2 \
    --tool-call-parser mistral --enable-auto-tool-choice

1. To ping the client you can use a simple Python snippet.

import requests
import json
from huggingface_hub import hf_hub_download


url = "http://<your-server-url>:8000/v1/chat/completions"
headers = {"Content-Type": "application/json", "Authorization": "Bearer token"}

model = "mistralai/Devstral-Small-2-24B-Instruct-2512"

def load_system_prompt(repo_id: str, filename: str) -> str:
    file_path = hf_hub_download(repo_id=repo_id, filename=filename)
    with open(file_path, "r") as file:
        system_prompt = file.read()
    return system_prompt

SYSTEM_PROMPT = load_system_prompt(model, "CHAT_SYSTEM_PROMPT.txt")

messages = [
    {"role": "system", "content": SYSTEM_PROMPT},
    {
        "role": "user",
        "content": [
            {
                "type": "text",
                "text": "<your-command>",
            },
        ],
    },
]

data = {"model": model, "messages": messages, "temperature": 0.15}

# Devstral Small 2 supports tool calling. If you want to use tools, follow this:
# tools = [ # Define tools for vLLM
#     {
#         "type": "function",
#         "function": {
#             "name": "git_clone",
#             "description": "Clone a git repository",
#             "parameters": {
#                 "type": "object",
#                 "properties": {
#                     "url": {
#                         "type": "string",
#                         "description": "The url of the git repository",
#                     },
#                 },
#                 "required": ["url"],
#             },
#         },
#     }
# ] 
# data = {"model": model, "messages": messages, "temperature": 0.15, "tools": tools} # Pass tools to payload.

response = requests.post(url, headers=headers, data=json.dumps(data))
print(response.json()["choices"][0]["message"]["content"])

SGLang

<details>
<summary>Expand</summary>

To use this model with SGLang to implement a production-ready inference pipelines (OpenAI-compatible API server),
see the following sections.

Install SGLang from source (track latest main locally):

git clone https://github.com/sgl-project/sglang.git
cd sglang
uv pip install -e python
uv pip install transformers==5.0.0rc # required
uv pip install nvidia-cudnn-cu12==9.16.0.29 # required for VLM

We recommend that you use Devstral Small 2 in a server/client setting.

1. Spin up a server:

python -m sglang.launch_server --model-path mistralai/Devstral-Small-2-24B-Instruct-2512 --host 0.0.0.0 --port 30000 --tp 2 --tool-call-parser mistral

1. To ping the client you can use a simple Python snippet.

import requests
import json
from huggingface_hub import hf_hub_download


url = "http://<your-server-url>:30000/v1/chat/completions"
headers = {"Content-Type": "application/json", "Authorization": "Bearer token"}

model = "mistralai/Devstral-Small-2-24B-Instruct-2512"

def load_system_prompt(repo_id: str, filename: str) -> str:
    file_path = hf_hub_download(repo_id=repo_id, filename=filename)
    with open(file_path, "r") as file:
        system_prompt = file.read()
    return system_prompt

SYSTEM_PROMPT = load_system_prompt(model, "CHAT_SYSTEM_PROMPT.txt")

messages = [
    {"role": "system", "content": SYSTEM_PROMPT},
    {
        "role": "user",
        "content": [
            {
                "type": "text",
                "text": "<your-command>",
            },
        ],
    },
]

data = {"model": model, "messages": messages, "temperature": 0.15}

# Devstral Small 2 supports tool calling. If you want to use tools, follow this:
# tools = [  # Define tools (OpenAI-compatible)
#     {
#         "type": "function",
#         "function": {
#             "name": "git_clone",
#             "description": "Clone a git repository",
#             "parameters": {
#                 "type": "object",
#                 "properties": {
#                     "url": {
#                         "type": "string",
#                         "description": "The url of the git repository",
#                     },
#                 },
#                 "required": ["url"],
#             },
#         },
#     }
# ] 
# data = {"model": model, "messages": messages, "temperature": 0.15, "tools": tools} # Pass tools to payload.

response = requests.post(url, headers=headers, data=json.dumps(data))
print(response.json()["choices"][0]["message"]["content"])

Transformers

<details>
<summary>Expand</summary

Make sure to install from main:

uv pip install git+https://github.com/huggingface/transformers

And run the following code snippet:

import torch
from transformers import (
    Mistral3ForConditionalGeneration,
    MistralCommonBackend,
)

model_id = "mistralai/Devstral-Small-2-24B-Instruct-2512"

tokenizer = MistralCommonBackend.from_pretrained(model_id)
model = Mistral3ForConditionalGeneration.from_pretrained(model_id, device_map="auto")

SP = """You are operating as and within Mistral Vibe, a CLI coding-agent built by Mistral AI and powered by default by the Devstral family of models. It wraps Mistral's Devstral models to enable natural language interaction with a local codebase. Use the available tools when helpful.

You can:

- Receive user prompts, project context, and files.
- Send responses and emit function calls (e.g., shell commands, code edits).
- Apply patches, run commands, based on user approvals.

Answer the user's request using the relevant tool(s), if they are available. Check that all the required parameters for each tool call are provided or can reasonably be inferred from context. IF there are no relevant tools or there are missing values for required parameters, ask the user to supply these values; otherwise proceed with the tool calls. If the user provides a specific value for a parameter (for example provided in quotes), make sure to use that value EXACTLY. DO NOT make up values for or ask about optional parameters. Carefully analyze descriptive terms in the request as they may indicate required parameter values that should be included even if not explicitly quoted.

Always try your hardest to use the tools to answer the user's request. If you can't use the tools, explain why and ask the user for more information.

Act as an agentic assistant, if a user asks for a long task, break it down and do it step by step.

When you want to commit changes, you will always use the 'git commit' bash command. It will always be suffixed with a line telling it was generated by Mistral Vibe with the appropriate co-authoring information. The format you will always use is the following heredoc.

Generated by Mistral Vibe.
Co-Authored-By: Mistral Vibe <vibe@mistral.ai>"
INLINECODE13zBF2DA287th element with n a parameter passed to the function ? You should start the sequence from 1, previous values are invalid.\nThen run the Python code for the function for n=5 and give the answer.", } ], }, ], "tools": [ { "type": "function", "function": { "name": "add_number", "description": "Add two numbers.", "parameters": { "type": "object", "properties": { "a": {"type": "string", "description": "The first number."}, "b": {"type": "string", "description": "The second number."}, }, "required": ["a", "b"], }, }, }, { "type": "function", "function": { "name": "multiply_number", "description": "Multiply two numbers.", "parameters": { "type": "object", "properties": { "a": {"type": "string", "description": "The first number."}, "b": {"type": "string", "description": "The second number."}, }, "required": ["a", "b"], }, }, }, { "type": "function", "function": { "name": "substract_number", "description": "Substract two numbers.", "parameters": { "type": "object", "properties": { "a": {"type": "string", "description": "The first number."}, "b": {"type": "string", "description": "The second number."}, }, "required": ["a", "b"], }, }, }, { "type": "function", "function": { "name": "write_a_story", "description": "Write a story about science fiction and people with badass laser sabers.", "parameters": {}, }, }, { "type": "function", "function": { "name": "terminal", "description": "Perform operations from the terminal.", "parameters": { "type": "object", "properties": { "command": { "type": "string", "description": "The command you wish to launch, e.g ls, rm, ...", }, "args": { "type": "string", "description": "The arguments to pass to the command.", }, }, "required": ["command"], }, }, }, { "type": "function", "function": { "name": "python", "description": "Call a Python interpreter with some Python code that will be ran.", "parameters": { "type": "object", "properties": { "code": { "type": "string", "description": "The Python code to run", }, "result_variable": { "type": "string", "description": "Variable containing the result you'd like to retrieve from the execution.", }, }, "required": ["code", "result_variable"], }, }, }, ], } tokenized = tokenizer.apply_chat_template( conversation=input["messages"], tools=input["tools"], return_tensors="pt", return_dict=True, ) input_ids = tokenized["input_ids"].to(device="cuda") output = model.generate( input_ids, max_new_tokens=200, do_sample=True, temperature=0.15, )[0] decoded_output = tokenizer.decode(output[len(tokenized["input_ids"][0]) :]) print(decoded_output)

</details>

## Tests
To help test our model via vLLM or test that other frameworks' implementations are correct, here is a set of prompts you can try with the expected outputs.

1. Call one tool 

<details>
  <summary>Messages and tools</summary>

</details>

<details>
  <summary>Expected output trace</summary>

</details>

2. Call tools one at a time subsequently

<details>
  <summary>Messages and tools</summary>

</details>

<details>
  <summary>Expected output trace</summary>