Supertron2-24B: A Capable Instruction-Tuned Coding and Reasoning Model
Model Description
Supertron2-24B is an instruction-tuned language model built on top of mistralai/Devstral-Small-2-24B-Instruct-2512. It is designed for practical coding assistance, structured reasoning, math, science, general chat, and everyday instruction following.
- Developed by: Surpem
- Model type: Causal Language Model
- Architecture: Dense Transformer, 24B parameters
- License: Apache 2.0
Capabilities
Coding
Supertron2-24B is designed to help write, explain, and debug code. It can assist with practical programming tasks, implementation planning, error analysis, and code review style explanations.
Reasoning
The model can work through multi-step questions, compare options, follow structured instructions, and produce concise answers when requested.
Math
Supertron2-24B can handle arithmetic, algebra-style problems, word problems, and step-by-step mathematical explanations.
Science
The model can explain scientific concepts clearly, answer STEM questions, and help with educational or technical writing.
General Chat
Supertron2-24B can assist with writing, brainstorming, explanations, planning, summarization, and general everyday questions.
Get Started
from transformers import AutoTokenizer, AutoModelForImageTextToText
import torch
model_id = "Surpem/Supertron2-24B"
tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
model = AutoModelForImageTextToText.from_pretrained(
model_id,
torch_dtype=torch.bfloat16,
device_map="auto",
trust_remote_code=True,
)
messages = [
{"role": "user", "content": "Write a Python function that checks if a string is a palindrome."}
]
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
| Precision | Min VRAM | Recommended |
|---|---|---|
| bfloat16 | 48 GB | 80 GB+ |
| 4-bit quantized | 16 GB | 24 GB+ |
For long contexts or larger batches, use more VRAM or reduce batch size and max sequence length.
Intended Use
Supertron2-24B is intended for:
- Coding assistance
- Software engineering reasoning
- Math and science explanations
- General chat and instruction following
- Writing, summarization, and brainstorming
- Research and technical assistance
Limitations
- The model can make mistakes and should be checked for important work.
- It may produce incorrect code, incomplete reasoning, or outdated information.
- It should not be used as the only source for medical, legal, financial, or safety-critical decisions.
- Generated code should be reviewed and tested before use.
Citation
@misc{surpem2026supertron2-24b,
title={Supertron2-24B -- Instruction-Tuned Coding and Reasoning Model},
author={Surpem},
year={2026},
url={https://huggingface.co/Surpem/Supertron2-24B},
}