---

language:

• en
• fr
• es
• de
• it
• pt
• nl
• zh
• ja
• ko
• ar

license: apache-2.0

inference: false

base_model:

• mistralai/Ministral-3-14B-Base-2512

extra_gated_description: >-

If you want to learn more about how we process your personal data, please read

our <a href="https://mistral.ai/terms/">Privacy Policy</a>.

tags:

• mistral-common
• RSIT
• Infiniteons

---

Quantized from mistralai/Ministral-3-14B-Instruct-2512 (BF16) using the Infiniteon Algebra associator to guide mixed-precision allocation: 20% of the weights (the shared expert and attention tensors) are quantized to 4-bit (Q4_K), and the remaining 80% to 3-bit (Q3_K). This achieves Q4-level generation quality at near-Q3 size, fitting a 14B-parameter model inside 8 GB VRAM.

Why this works: The associator of the Infiniteon algebra identifies which weight tensors carry structural "gap-crossing" information — the tensors that connect diverse token neighborhoods. By allocating 4-bit precision to these structurally critical tensors and 3-bit to the rest, the model preserves the directional diversity needed for coherent generation while keeping the overall size near Q3.

Research and credits: This research was conducted by TNT.Chat (https://ai.tnt.chat) as part of our LOCAL AI effort to bring larger models onto consumer hardware. The target is a capable model running on 8 GB VRAM for Agentic usage with a large context window. Additional models targeting RTX 5090 and RTX 6000 PRO (full GLM-4.7 218B REAP) are in progress.

Theory references:

Pinto Martins, C. F. (2026). Ramsey Statistics And Infiniteons Theory, Volume I. Zenodo. DOI: https://doi.org/10.5281/zenodo.19329589

Pinto Martins, C. F. (2026). Ramsey Statistics And Infiniteons Theory, Volume II. Zenodo. DOI: https://doi.org/10.5281/zenodo.19330373

The mmproj GGUF has BF16 and Q3Q4 formats. The Q2Q4 format follows the same 20% Q4 and 80% Q2 structure.

---

Ministral 3 14B Instruct 2512

The largest model in the Ministral 3 family, Ministral 3 14B offers frontier capabilities and performance comparable to its larger Mistral Small 3.2 24B counterpart. A powerful and efficient language model with vision capabilities.

This model is the instruct post-trained version in FP8, fine-tuned for instruction tasks, making it ideal for chat and instruction based use cases.

The Ministral 3 family is designed for edge deployment, capable of running on a wide range of hardware. Ministral 3 14B can even be deployed locally, capable of fitting in 24GB of VRAM in FP8, and less if further quantized.

Learn more in our blog post and paper.

Key Features

Ministral 3 14B consists of two main architectural components:

• 13.5B Language Model
• 0.4B Vision Encoder

The Ministral 3 14B Instruct model offers the following capabilities:

• Vision: Enables the model to analyze images and provide insights based on visual content, in addition to text.
• Multilingual: Supports dozens of languages, including English, French, Spanish, German, Italian, Portuguese, Dutch, Chinese, Japanese, Korean, Arabic.
• System Prompt: Maintains strong adherence and support for system prompts.
• Agentic: Offers best-in-class agentic capabilities with native function calling and JSON outputting.
• Edge-Optimized: Delivers best-in-class performance at a small scale, deployable anywhere.
• Apache 2.0 License: Open-source license allowing usage and modification for both commercial and non-commercial purposes.
• Large Context Window: Supports a 256k context window.

Use Cases

Private AI deployments where advanced capabilities meet practical hardware constraints:

• Private/custom chat and AI assistant deployments in constrained environments
• Advanced local agentic use cases
• Fine-tuning and specialization
• And more...

Bringing advanced AI capabilities to most environments.

Recommended Settings

We recommend deploying with the following best practices:

• System Prompt: Define a clear environment and use case, including guidance on how to effectively leverage tools in agentic systems.
• Sampling Parameters: Use a temperature below 0.1 for daily-driver and production environments ; Higher temperatures may be explored for creative use cases - developers are encouraged to experiment with alternative settings.
• Tools: Keep the set of tools well-defined and limit their number to the minimum required for the use case - Avoiding overloading the model with an excessive number of tools.
• Vision: When deploying with vision capabilities, we recommend maintaining an aspect ratio close to 1:1 (width-to-height) for images. Avoiding the use of overly thin or wide images - crop them as needed to ensure optimal performance.

Ministral 3 Family

| Model Name | Type | Precision | Link |

|--------------------------------|--------------------|-----------|------------------------------------------------------------------------------------------|

| Ministral 3 3B Base 2512 | Base pre-trained | BF16 | Hugging Face |

| Ministral 3 3B Instruct 2512 | Instruct post-trained | FP8 | Hugging Face |

| Ministral 3 3B Reasoning 2512 | Reasoning capable | BF16 | Hugging Face |

| Ministral 3 8B Base 2512 | Base pre-trained | BF16 | Hugging Face |

| Ministral 3 8B Instruct 2512 | Instruct post-trained | FP8 | Hugging Face |

| Ministral 3 8B Reasoning 2512 | Reasoning capable | BF16 | Hugging Face |

| Ministral 3 14B Base 2512 | Base pre-trained | BF16 | Hugging Face |

| Ministral 3 14B Instruct 2512 | Instruct post-trained | FP8 | Hugging Face |

| Ministral 3 14B Reasoning 2512 | Reasoning capable | BF16 | Hugging Face |

Other formats available here.

Benchmark Results

We compare Ministral 3 to similar sized models.

Reasoning

| Model | AIME25 | AIME24 | GPQA Diamond | LiveCodeBench |

|---------------------------|-------------|-------------|--------------|---------------|

| Ministral 3 14B | <u>0.850</u>| <u>0.898</u>| <u>0.712</u> | <u>0.646</u> |

| Qwen3-14B (Thinking) | 0.737 | 0.837 | 0.663 | 0.593 |

| | | | | |

| Ministral 3 8B | 0.787 | <u>0.860</u>| 0.668 | <u>0.616</u> |

| Qwen3-VL-8B-Thinking | <u>0.798</u>| <u>0.860</u>| <u>0.671</u> | 0.580 |

| | | | | |

| Ministral 3 3B | <u>0.721</u>| <u>0.775</u>| 0.534 | <u>0.548</u> |

| Qwen3-VL-4B-Thinking | 0.697 | 0.729 | <u>0.601</u> | 0.513 |

Instruct

| Model | Arena Hard | WildBench | MATH Maj@1 | MM MTBench |

|---------------------------|-------------|------------|-------------|------------------|

| Ministral 3 14B | <u>0.551</u>| <u>68.5</u>| <u>0.904</u>| <u>8.49</u> |

| Qwen3 14B (Non-Thinking) | 0.427 | 65.1 | 0.870 | NOT MULTIMODAL |

| Gemma3-12B-Instruct | 0.436 | 63.2 | 0.854 | 6.70 |

| | | | | |

| Ministral 3 8B | 0.509 | <u>66.8</u>| 0.876 | <u>8.08</u> |

| Qwen3-VL-8B-Instruct | <u>0.528</u>| 66.3 | <u>0.946</u>| 8.00 |

| | | | | |

| Ministral 3 3B | 0.305 | <u>56.8</u>| 0.830 | 7.83 |

| Qwen3-VL-4B-Instruct | <u>0.438</u>| <u>56.8</u>| <u>0.900</u>| <u>8.01</u> |

| Qwen3-VL-2B-Instruct | 0.163 | 42.2 | 0.786 | 6.36 |

| Gemma3-4B-Instruct | 0.318 | 49.1 | 0.759 | 5.23 |

Base

| Model | Multilingual MMLU | MATH CoT 2-Shot | AGIEval 5-shot | MMLU Redux 5-shot | MMLU 5-shot | TriviaQA 5-shot |

|---------------------|-------------------|-----------------|----------------|-------------------|-------------|-----------------|

| Ministral 3 14B | 0.742 | <u>0.676</u> | 0.648 | 0.820 | 0.794 | 0.749 |

| Qwen3 14B Base | <u>0.754</u> | 0.620 | <u>0.661</u> | <u>0.837</u> | <u>0.804</u>| 0.703 |

| Gemma 3 12B Base | 0.690 | 0.487 | 0.587 | 0.766 | 0.745 | <u>0.788</u> |

| | | | | | | |

| Ministral 3 8B | <u>0.706</u> | <u>0.626</u> | 0.591 | 0.793 | <u>0.761</u>| <u>0.681</u> |

| Qwen 3 8B Base | 0.700 | 0.576 | <u>0.596</u> | <u>0.794</u> | 0.760 | 0.639 |

| | | | | | | |

| Ministral 3 3B | 0.652 | <u>0.601</u> | 0.511 | 0.735 | 0.707 | 0.592 |

| Qwen 3 4B Base | <u>0.677</u> | 0.405 | <u>0.570</u> | <u>0.759</u> | <u>0.713</u>| 0.530 |

| Gemma 3 4B Base | 0.516 | 0.294 | 0.430 | 0.626 | 0.589 | <u>0.640</u> |

Usage

The model can be used with the following frameworks;

• vllm: See here
• transformers: See here

vLLM

We recommend using this model with vLLM.

Installation

Make sure to install vllm >= 0.12.0:

pip install vllm --upgrade

Doing so should automatically install mistral_common >= 1.8.6.

To check:

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

You can also make use of a ready-to-go docker image or on the docker hub.

Serve

Due to their size and the FP8 format of their weights Ministral-3-3B-Instruct-2512, Ministral-3-8B-Instruct-2512 and Ministral-3-14B-Instruct-2512 can run on a single 1xH200 GPU.

A simple launch command is:

vllm serve mistralai/Ministral-3-14B-Instruct-2512 \
  --tokenizer_mode mistral --config_format mistral --load_format mistral \
  --enable-auto-tool-choice --tool-call-parser mistral

Key parameter notes:

• enable-auto-tool-choice: Required when enabling tool usage.
• tool-call-parser mistral: Required when enabling tool usage.

Additional flags:

• You can set --max-model-len to preserve memory. By default it is set to 262144 which is quite large but not necessary for most scenarios.
• You can set --max-num-batched-tokens to balance throughput and latency, higher means higher throughput but higher latency.

Usage of the model

Here we assume that the model mistralai/Ministral-3-14B-Instruct-2512 is served and you can ping it to the domain localhost with the port 8000 which is the default for vLLM.

<details>

<summary>Vision Reasoning</summary>

Let's see if the Ministral 3 knows when to pick a fight !

from datetime import datetime, timedelta

from openai import OpenAI
from huggingface_hub import hf_hub_download

# Modify OpenAI's API key and API base to use vLLM's API server.
openai_api_key = "EMPTY"
openai_api_base = "http://localhost:8000/v1"

TEMP = 0.15
MAX_TOK = 262144

client = OpenAI(
    api_key=openai_api_key,
    base_url=openai_api_base,
)

models = client.models.list()
model = models.data[0].id


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()
    today = datetime.today().strftime("%Y-%m-%d")
    yesterday = (datetime.today() - timedelta(days=1)).strftime("%Y-%m-%d")
    model_name = repo_id.split("/")[-1]
    return system_prompt.format(name=model_name, today=today, yesterday=yesterday)


SYSTEM_PROMPT = load_system_prompt(model, "SYSTEM_PROMPT.txt")
image_url = "https://static.wikia.nocookie.net/essentialsdocs/images/7/70/Battle.png/revision/latest?cb=20220523172438"

messages = [
    {"role": "system", "content": SYSTEM_PROMPT},
    {
        "role": "user",
        "content": [
            {
                "type": "text",
                "text": "What action do you think I should take in this situation? List all the possible actions and explain why you think they are good or bad.",
            },
            {"type": "image_url", "image_url": {"url": image_url}},
        ],
    },
]


response = client.chat.completions.create(
    model=model,
    messages=messages,
    temperature=TEMP,
    max_tokens=MAX_TOK,
)

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

</details>

<details>

<summary>Function Calling</summary>

Let's solve some equations thanks to our simple Python calculator tool.

import json
from openai import OpenAI
from huggingface_hub import hf_hub_download

# Modify OpenAI's API key and API base to use vLLM's API server.
openai_api_key = "EMPTY"
openai_api_base = "http://localhost:8000/v1"

TEMP = 0.15
MAX_TOK = 262144

client = OpenAI(
    api_key=openai_api_key,
    base_url=openai_api_base,
)

models = client.models.list()
model = models.data[0].id


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, "SYSTEM_PROMPT.txt")

image_url = "https://math-coaching.com/img/fiche/46/expressions-mathematiques.jpg"


def my_calculator(expression: str) -> str:
    return str(eval(expression))


tools = [
    {
        "type": "function",
        "function": {
            "name": "my_calculator",
            "description": "A calculator that can evaluate a mathematical expression.",
            "parameters": {
                "type": "object",
                "properties": {
                    "expression": {
                        "type": "string",
                        "description": "The mathematical expression to evaluate.",
                    },
                },
                "required": ["expression"],
            },
        },
    },
    {
        "type": "function",
        "function": {
            "name": "rewrite",
            "description": "Rewrite a given text for improved clarity",
            "parameters": {
                "type": "object",
                "properties": {
                    "text": {
                        "type": "string",
                        "description": "The input text to rewrite",
                    }
                },
            },
        },
    },
]

messages = [
    {"role": "system", "content": SYSTEM_PROMPT},
    {
        "role": "user",
        "content": [
            {
                "type": "text",
                "text": "Thanks to your calculator, compute the results for the equations that involve numbers displayed in the image.",
            },
            {
                "type": "image_url",
                "image_url": {
                    "url": image_url,
                },
            },
        ],
    },
]

response = client.chat.completions.create(
    model=model,
    messages=messages,
    temperature=TEMP,
    max_tokens=MAX_TOK,
    tools=tools,
    tool_choice="auto",
)

tool_calls = response.choices[0].message.tool_calls

results = []
for tool_call in tool_calls:
    function_name = tool_call.function.name
    function_args = tool_call.function.arguments
    if function_name == "my_calculator":
        result = my_calculator(**json.loads(function_args))
        results.append(result)

messages.append({"role": "assistant", "tool_calls": tool_calls})
for tool_call, result in zip(tool_calls, results):
    messages.append(
        {
            "role": "tool",
            "tool_call_id": tool_call.id,
            "name": tool_call.function.name,
            "content": result,
        }
    )


response = client.chat.completions.create(
    model=model,
    messages=messages,
    temperature=TEMP,
    max_tokens=MAX_TOK,
)

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

</details>

<details>

<summary>Text-Only Request</summary>

Ministral 3 can follow your instructions to the letter.

from openai import OpenAI
from huggingface_hub import hf_hub_download

# Modify OpenAI's API key and API base to use vLLM's API server.
openai_api_key = "EMPTY"
openai_api_base = "http://localhost:8000/v1"

TEMP = 0.15
MAX_TOK = 262144

client = OpenAI(
    api_key=openai_api_key,
    base_url=openai_api_base,
)

models = client.models.list()
model = models.data[0].id


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, "SYSTEM_PROMPT.txt")

messages = [
    {"role": "system", "content": SYSTEM_PROMPT},
    {
        "role": "user",
        "content": "Write me a sentence where every word starts with the next letter in the alphabet - start with 'a' and end with 'z'.",
    },
]

response = client.chat.completions.create(
    model=model,
    messages=messages,
    temperature=TEMP,
    max_tokens=MAX_TOK,
)

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

</details>

Transformers

You can also use Ministral 3 14B Instruct 2512 with Transformers !

Transformers recently added support for FP8, so make sure to install from main:

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

To make the best use of our model with Transformers make sure to have installed mistral-common >= 1.8.6 to use our tokenizer.

pip install mistral-common --upgrade

Try it out by running the following snippet.

> [!Tip]

> On latest main as of 05/12/2025, by default

> a FP8 triton kernel for fast accelerated matmuls

> (w8a8_block_fp8_matmul_triton) will be used

> without any degradation in accuracy. However, if you want to

> run your model in BF16 see (here)

<details>

<summary>Python snippet</summary>

import torch
from transformers import Mistral3ForConditionalGeneration, MistralCommonBackend

model_id = "mistralai/Ministral-3-14B-Instruct-2512"

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

image_url = "https://static.wikia.nocookie.net/essentialsdocs/images/7/70/Battle.png/revision/latest?cb=20220523172438"

messages = [
    {
        "role": "user",
        "content": [
            {
                "type": "text",
                "text": "What action do you think I should take in this situation? List all the possible actions and explain why you think they are good or bad.",
            },
            {"type": "image_url", "image_url": {"url": image_url}},
        ],
    },
]

tokenized = tokenizer.apply_chat_template(messages, return_tensors="pt", return_dict=True)

tokenized["input_ids"] = tokenized["input_ids"].to(device="cuda")
tokenized["pixel_values"] = tokenized["pixel_values"].to(dtype=torch.bfloat16, device="cuda")
image_sizes = [tokenized["pixel_values"].shape[-2:]]

output = model.generate(
    **tokenized,
    image_sizes=image_sizes,
    max_new_tokens=512,
)[0]

decoded_output = tokenizer.decode(output[len(tokenized["input_ids"][0]):])
print(decoded_output)

</details>

Transformers BF16

Transformers allows you to automatically convert the checkpoint to Bfloat16. To do so, simply load the model as follows:

from transformers import Mistral3ForConditionalGeneration, FineGrainedFP8Config

model_id = "mistralai/Ministral-3-14B-Instruct-2512"
model = Mistral3ForConditionalGeneration.from_pretrained(
    model_id,
    device_map="auto",
    quantization_config=FineGrainedFP8Config(dequantize=True)
)

License

This model is licensed under the Apache 2.0 License.

You must not use this model in a manner that infringes, misappropriates, or otherwise violates any third party’s rights, including intellectual property rights.