🧠 Mistral-Small-3.1-24B-Instruct-2503-Ko-Reasoning

A large-scale Korean reasoning model fine-tuned from mistralai/Mistral-Small-3.1-24B-Instruct-2503, designed to excel in logical and multi-hop reasoning tasks in Korean.

---

📌 Overview

Mistral-Small-3.1-24B-Instruct-2503-Ko-Reasoning is a fine-tuned version of mistralai/Mistral-Small-3.1-24B-Instruct-2503, specifically optimized for logical reasoning in Korean. This model is part of a broader research initiative to explore:

• The transition from multilingual reasoning LLMs to Korean-specialized reasoning models
• The enhancement of non-reasoning Korean language models into reasoning-capable variants
• The development of open-access models that rival proprietary alternatives in complex reasoning tasks

This model was fine-tuned using a large-scale Korean-English instruction dataset containing diverse multi-hop questions, symbolic logic tasks, and human-crafted reasoning steps.

---

🧪 Benchmark Results

• 📊 All benchmarks were measured using the 0-shot CoT (Chain-of-Thought) method.
• 📊 The Score represents either the accuracy (%) of correct answers or a rating on a 1-10 scale from a judge model.
• 📊 LLM-as-a-judge benchmarks were evaluated using GPT-4o (2024-08-01-preview).

Benchmark	Score
GPQA diamond	79.3
GSM8K	68.7
HAERAE	93.3
KSM	80.9
LogicKor	8.52
Math500	76.2
MT-Bench	8.29
MT-Bench(Ko)	8.03

---

🧑‍💻 Usage

Install Transformers >= 4.50:

pip install -U transformers

Basic example:

from transformers import AutoProcessor, Mistral3ForConditionalGeneration
from PIL import Image
import requests
import torch

model_id = "DimensionSTP/Mistral-Small-3.1-24B-Instruct-2503-Ko-Reasoning"

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

processor = AutoProcessor.from_pretrained(model_id)

messages = [
    {
        "role": "system",
        "content": [{"type": "text", "text": "You are a helpful assistant."}]
    },
    {
        "role": "user",
        "content": [
            {"type": "text", "text": "서울과 부산 중 어디가 더 커?"}
        ]
    }
]

inputs = processor.apply_chat_template(
    messages, add_generation_prompt=True, tokenize=True,
    return_dict=True, return_tensors="pt"
).to(model.device, dtype=torch.bfloat16)

input_len = inputs["input_ids"].shape[-1]

with torch.inference_mode():
    generation = model.generate(**inputs, max_new_tokens=8192, do_sample=False)
    generation = generation[0][input_len:]

decoded = processor.decode(generation, skip_special_tokens=True)
print(decoded)

---

🧠 Base Model: mistralai/Mistral-Small-3.1-24B-Instruct-2503

The base model, mistralai/Mistral-Small-3.1-24B-Instruct-2503, is a VLM developed by the Mistral team. For more technical details, refer to the Mistral Small 3.1 News.

---

🧱 Model Architecture

Property	Value
Architecture	Mistral3ForConditionalGeneration
Parameters	24B
Context Length	131,072 tokens
Tokenizer	LlamaTokenizer (BPE)

---

📅 Release Date

Mar 2025
This model was released in March 2025 as part of the Ko-Reasoning Series, which focuses on pushing the boundaries of open-source reasoning in Korean using modern LLMs.

---

📬 Contact

For questions, collaborations, or deployment inquiries, please contact:

• 🤖 Hugging Face: https://huggingface.co/DimensionSTP
• ✉️ Email: [[email protected]]

---

📦 Available Checkpoints

• ✅ main: Final stable version from the last branch
• ✅ All training artifacts available (tokenizer, config, model weights)