README.md

---
language:
- tr
- en
- de
- es
- fr
- ru
- zh
- ja
- ko
license: mit
tags:
- turkish
- türkiye
- reasoning
- ai
- lamapi
- gemma3
- next
- next-x1
- text-generation
- open-source
- 14b
- large-language-model
- llm
- transformer
- artificial-intelligence
- machine-learning
- nlp
- multilingual
- instruction-tuned
- chat
- generative-ai
- optimized
- trl
- sft
- cognitive
- analytical
- enterprise
pipeline_tag: text-generation
datasets:
- CognitiveKernel/CognitiveKernel-Pro-SFT
- OpenSPG/KAG-Thinker-training-dataset
- QuixiAI/dolphin-r1
- uclanlp/Brief-Pro
- Gryphe/Opus-WritingPrompts
- GreenerPastures/All-Your-Base-Full
- dongguanting/ARPO-SFT-54K
- Medint/Multi-Med-conversational
- mlabonne/smoltalk-flat
- mlabonne/natural_reasoning-formatted
- QuixiAI/open-instruct-uncensored
- mlabonne/open-perfectblend
library_name: transformers
---

<img src='assets/banner.png'>

# 🧠 Next 8B (m427)

### *Türkiye’s Compact Reasoning AI — Logical, Analytical, and Efficient*

[![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](https://opensource.org/licenses/MIT)
[![Language: Multilingual](https://img.shields.io/badge/Language-Multilingual-red.svg)]()
[![HuggingFace](https://img.shields.io/badge/🤗-Lamapi/Next--8B-orange.svg)](https://huggingface.co/Lamapi/next-8b)

---

## 📖 Overview

**Next 8B** is an **8-billion parameter large language model (LLM)** built on **Qwen 3 architecture**, optimized for **reasoning and analytical performance**.
It’s **Türkiye’s reasoning-capable compact AI**, designed to think, infer, and solve problems efficiently.

Focused purely on **cognitive tasks**, it excels in problem-solving, abstract logic, and multilingual understanding (Turkish, English, and more).

---

## ⚡ Highlights

* 🇹🇷 **Türkiye’s compact reasoning AI**
* 🧠 **Logical, analytical, and inferential reasoning**
* 🌍 **Multilingual support (Turkish, English, 30+ languages)**
* ⚡ **Lightweight and efficient**
* 💬 **Instruction-tuned for dialogue, tutoring, and analysis**

---

## 📊 Benchmark Performance

<table>
  <thead>
    <tr>
      <th>Model</th>
      <th>MMLU (5-shot) %</th>
      <th>MMLU-Pro %</th>
      <th>GSM8K %</th>
      <th>MATH %</th>
    </tr>
  </thead>
  <tbody>
    <tr>
      <td>Next 14B (Thinking)</td>
      <td><strong>94.6</strong></td>
      <td><strong>93.2</strong></td>
      <td><strong>98.8</strong></td>
      <td>92.7</td>
    </tr>
    <tr>
      <td>Next 12B</td>
      <td>92.7</td>
      <td>84.4</td>
      <td>95.3</td>
      <td>87.2</td>
    </tr>
    <tr class="next">
      <td><strong>Next 8B (Thinking)</strong></td>
      <td>91.0</td>
      <td>88.5</td>
      <td>96.2</td>
      <td>88.0</td>
    </tr>
    <tr>
      <td>GPT-5</td>
      <td>92.5</td>
      <td>87.0</td>
      <td>98.4</td>
      <td><strong>96.0</strong></td>
    </tr>
    <tr>
      <td>Claude Opus 4.1 (Thinking)</td>
      <td>~92.0</td>
      <td>87.8</td>
      <td>84.7</td>
      <td>95.4</td>
    </tr>
  </tbody>
</table>



---

## 🚀 Installation & Usage

```python
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

model_id = "Lamapi/next-8b"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.float16, device_map="auto")

messages = [
    {"role": "system", "content": "You are Next-X1, a reasoning-capable AI assistant created by Lamapi. You think logically, reason efficiently, and answer concisely."},
    {"role": "user", "content": "Explain why the sky appears blue using logical reasoning."}
]

prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

outputs = model.generate(**inputs, max_new_tokens=150)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
```

---

## 🧩 Key Features

| Feature                                | Description                                                                  |
| -------------------------------------- | ---------------------------------------------------------------------------- |
| 🧠 **Efficient Reasoning**             | Strong in abstract logic, critical thinking, and structured problem-solving. |
| 🇹🇷 **Multilingual Intelligence**     | Deep Turkish understanding with 30+ language support.                        |
| ⚡ **Lightweight & Optimized**          | Quantized formats (Q8_0, Q4_K_M, FP16) for efficiency.                       |
| 🧮 **Mathematical & Analytical Skill** | Handles structured reasoning and moderate complexity problems.               |
| 🧩 **Non-Vision Architecture**         | Focused on text-based cognitive tasks.                                       |
| 🏢 **Reliable & Consistent**           | Predictable outputs suitable for professional use.                           |

---

## 📐 Model Specifications

| Specification     | Details                                                       |
| ----------------- | ------------------------------------------------------------- |
| **Base Model**    | Qwen 3                                                        |
| **Parameters**    | 8 Billion                                                     |
| **Architecture**  | Transformer (Causal LLM)                                      |
| **Modalities**    | Text-only                                                     |
| **Fine-Tuning**   | Instruction-tuned with reasoning datasets                     |
| **Optimizations** | Quantization-ready, FP16 support                              |
| **Primary Focus** | Reasoning, logic, decision-making, and language understanding |

---

## 🎯 Ideal Use Cases

* **Compact Analytical Chatbots**
* **Research Assistance** (scientific/legal)
* **Education & Tutoring**
* **Code & Algorithm Design**
* **Decision Support Systems**

---

## 💡 Performance Highlights

* **Efficient Reasoning:** Compact yet powerful logical reasoning.
* **Good Mathematical Understanding:** Handles structured problems reliably.
* **Lightweight & Fast:** Ideal for resource-conscious environments.
* **Consistent Outputs:** Professional-grade reliability in smaller footprint.

---

## 📄 License

Licensed under **MIT License** — free for commercial and non-commercial use.

---

## 📞 Contact & Support

* 📧 **Email:** [[email protected]](mailto:[email protected])
* 🤗 **HuggingFace:** [Lamapi](https://huggingface.co/Lamapi)

---

> **Next 8B** — compact *reasoning-capable* AI, blending **logical depth**, **analytical efficiency**, and **lightweight reliability**.

[![Follow on HuggingFace](https://img.shields.io/badge/Follow-HuggingFace-yellow?logo=huggingface)](https://huggingface.co/Lamapi)