| Model | MMLU (5-shot) % | MMLU-Pro (Reasoning) % | GSM8K % | MATH % |
|---|---|---|---|---|
| Next 32B (Thinking) | 96.2 | 97.1 | 99.7 | 97.1 |
| GPT-5.1 | 98.4 | 95.9 | 99.7 | 98.5 |
| Claude Opus 4.5 | 97.5 | 96.5 | 99.2 | 97.8 |
| Gemini 3 Pro | 97.9 | 94.8 | 98.9 | 96.4 |
| Grok 4.1 | 96.1 | 92.4 | 97.8 | 95.2 |
| Next 14B (prev) | 94.6 | 93.2 | 98.8 | 92.7 |
!pip install unsloth1from unsloth import FastLanguageModel
2
3model, tokenizer = FastLanguageModel.from_pretrained("Lamapi/next-32b")
4
5messages = [
6 {"role": "system", "content": "You are Next-X1, an AI assistant created by Lamapi. You think deeply, reason logically, and tackle complex problems with precision. You are an helpful, smart, kind, concise AI assistant."},
7 {"role" : "user", "content" : "Analyze the potential long-term economic impacts of AI on emerging markets using a dialectical approach."}
8]
9text = tokenizer.apply_chat_template(
10 messages,
11 tokenize = False,
12 add_generation_prompt = True,
13 enable_thinking = True, # Enable thinking
14)
15
16from transformers import TextStreamer
17_ = model.generate(
18 **tokenizer(text, return_tensors = "pt").to("cuda"),
19 max_new_tokens = 1024, # Increase for longer outputs!
20 temperature = 0.7, top_p = 0.95, top_k = 400,
21 streamer = TextStreamer(tokenizer, skip_prompt = True),
22)| Feature | Description |
|---|---|
| 🧠 Deep Cognitive Architecture | Capable of handling massive context windows and multi-step logical chains. |
| 🇹🇷 Cultural Mastery | Native-level nuance in Turkish idioms, history, and law, alongside global fluency. |
| ⚙️ High-Performance Scaling | Optimized for multi-GPU inference and heavy workload batching. |
| 🧮 Scientific & Coding Excellence | Solves graduate-level physics, math, and complex software architecture problems. |
| 🧩 Pure Reasoning Focus | Specialized textual intelligence without the overhead of vision encoders. |
| 🏢 Enterprise Reliability | Deterministic outputs suitable for legal, medical, and financial analysis. |
| Specification | Details |
|---|---|
| Base Model | Qwen 3 |
| Parameters | 32 Billion |
| Architecture | Transformer (Causal LLM) |
| Modalities | Text-only |
| Fine-Tuning | Advanced SFT & RLHF on Cognitive Kernel & KAG-Thinker datasets |
| Optimizations | GQA, Flash Attention 3, Quantization-ready |
| Primary Focus | Deep Reasoning, Complex System Analysis, Strategic Planning |
Next 32B — Türkiye’s flagship reasoning model. Built for those who demand depth, precision, and massive intelligence.
