Supertron1-14B

Supertron1-14B: A Powerful Instruction-Tuned Language Model

Model Description

Supertron1-14B is a QLoRA fine-tuned language model built on top of Qwen3-14B. Trained with a focus on coding, mathematics, general knowledge, and safe reasoning, it delivers strong performance across technical and analytical tasks while maintaining a high standard of safety and helpfulness.

• Developed by: Surpem
• Model type: Causal Language Model
• Architecture: Dense Transformer, 14B parameters
• Fine-tuned from: Qwen/Qwen3-14B
• Fine-tuning method: QLoRA (4-bit NF4 + LoRA r=16, alpha=32, all-linear targets)
• License: Apache 2.0

---

Capabilities

Coding

Trained on tens of thousands of high-quality coding instruction pairs, the model can write, debug, explain, and refactor code across Python, JavaScript, C++, and more. It understands algorithmic thinking, software design patterns, and produces clean, well-commented output.

Mathematics

With dedicated training on competition-style math and step-by-step solutions, the model handles everything from algebra and calculus to advanced problem solving. It shows its full working rather than jumping to answers, making it reliable for both learning and verification.

General Knowledge

Broad instruction tuning across diverse domains means the model holds detailed technical conversations, assists with research and writing, explains difficult concepts clearly, and adapts to a wide range of tasks and formats.

Safety

Trained on the Anthropic HH-RLHF harmless dataset, the model is tuned to recognize and refuse harmful, illegal, or unethical requests with a brief, clear explanation. It prioritises helpfulness within safe boundaries.

Instruction Following

The model is highly responsive to natural language instructions and adapts its tone, format, and depth based on what you ask for — from concise one-liners to detailed technical walkthroughs.

---

Get Started

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

model_id = "Surpem/Supertron1-14B"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto"
)

messages = [
    {"role": "user", "content": "Write a Python function that checks if a number is prime."}
]

text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(text, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=512)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:], skip_special_tokens=True))

---

Hardware Requirements

For 4-bit quantized inference:

from transformers import BitsAndBytesConfig

bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype=torch.bfloat16,
    bnb_4bit_use_double_quant=True,
)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    quantization_config=bnb_config,
    device_map="auto"
)

---

Citation

@misc{surpem2026supertron1-14b,
      title={Supertron1-14B — Instruction-Tuned Language Model},
      author={Surpem},
      year={2026},
      url={https://huggingface.co/Surpem/Supertron1-14B},
}