Ministral-3-2512-ONNX

[!Tip] This model was contributed by Xenova from Hugging Face. We sincerely appreciate the integration and community collaboration. While preliminary functionality checks have been performed, comprehensive testing has not yet been completed. We recommend you to proceed with caution and conducting your own evaluations for specific use cases. If any issues arise, open a PR/Issue here and we will try to address them promptly.

Ministral 3 3B Instruct 2512 ONNX

The smallest model in the Ministral 3 family, Ministral 3 3B is a powerful, efficient tiny language model with vision capabilities.

This model quantized weights of the instruct post-trained version in ONNX, 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 3B can even be deployed locally, capable of fitting in 8GB of VRAM in FP8, and less if further quantized.

Learn more in our blog post and paper.

Key Features

Ministral 3 3B consists of two main architectural components:

• 3.4B Language Model
• 0.4B Vision Encoder

The Ministral 3 3B 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.

Usage

Transformers.js

Online WebGPU Demo

You can test out the model directly in your browser using our online WebGPU demo: https://huggingface.co/spaces/mistralai/Ministral_3B_WebGPU.

JavaScript

If you haven't already, you can install the Transformers.js JavaScript library from NPM using:

npm i @huggingface/transformers

You can then run the model as follows:

import {
    AutoProcessor,
    AutoModelForImageTextToText,
    load_image,
    TextStreamer,
} from "@huggingface/transformers";

// Load processor and model
const model_id = "mistralai/Ministral-3-3B-Instruct-2512-ONNX";
const processor = await AutoProcessor.from_pretrained(model_id);
const model = await AutoModelForImageTextToText.from_pretrained(model_id, {
    device: "webgpu",
});

// Prepare inputs
const messages = [
    {
        role: "user",
        content: [
            { type: "image" },
            {
                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.",
            },
        ],
    },
]
const prompt = processor.apply_chat_template(messages);
const url = "https://static.wikia.nocookie.net/essentialsdocs/images/7/70/Battle.png/revision/latest?cb=20220523172438";
const image = await load_image(url);
const inputs = await processor(image, prompt, { add_special_tokens: false });

// Generate response
const outputs = await model.generate({
    ...inputs,
    max_new_tokens: 2048,
    streamer: new TextStreamer(processor.tokenizer, {
        skip_prompt: true,
        // callback_function: (text) => { /* Do something with the streamed output */ },
    }),
});

// Decode output
const decoded = processor.batch_decode(
    outputs.slice(null, [inputs.input_ids.dims.at(-1), null]),
    { skip_special_tokens: true },
);
console.log(decoded[0]);

In this Pokémon game screenshot, you have several potential actions to consider, depending on the situation and your strategy. Here are the possible actions and their implications:

---

### **1. FIGHT**
**Description:** Use the Pikachu in your possession to attack the Pidgey.

#### **Pros:**
- **Potential to defeat Pidgey:** If Pikachu has strong moves (e.g., Thunderbolt, Electric Move) and Pidgey is weak to Electric-type attacks, this could be a strong choice.
- **Experience Gain:** Winning battles typically rewards you with experience points (XP) and sometimes drops Pidgey as a captured Pokémon.
- **Training Opportunity:** Helps you level up Pikachu, which can improve its stats and access new moves.

#### **Cons:**
- **Risk of Losing Pikachu:** If Pikachu is already at a low HP (e.g., 83/83 is not bad, but if it's lower), a loss could be detrimental.
- **Move Selection Matters:** If Pikachu doesn’t have a strong move against Pidgey, you might lose the battle unnecessarily.
- **Stamina Cost:** Battles can consume stamina if you're on a timer-based system (e.g., Pokémon Red/Blue).

---

### **2. RUN**
**Description:** Use the "Run" option to escape the battle.

#### **Pros:**
- **Avoids Risk:** If you're unsure about Pikachu’s moves or HP, running could save you from a potentially bad outcome.
- **Preserves Resources:** Running avoids losing Pikachu’s HP, which could be useful if you're running low.
- **Flexibility:** Allows you to explore or use other Pokémon later.

#### **Cons:**
- **No XP Gain:** Running means you won’t earn XP for defeating Pidgey.
- **Potential Consequences:** If you run too often, you might miss out on capturing Pidgey or gaining experience.
- **No Training:** Pikachu won’t level up or gain new moves if you avoid battles.

---

### **3. POKEMON (Select Another Pokémon)**
**Description:** If you have another Pokémon in your part