Dataset	Rows	Focus
`nohurry/Opus-4.6-Reasoning-3000x-filtered`	~2,330	Math, Coding
`TeichAI/Claude-Sonnet-4.6-Reasoning-1100x`	~1,096	Reasoning, Ethics, Economics, Policy
`TeichAI/Claude-Opus-4.6-Reasoning-887x`	~887	Bullshit detection, Legal, Vague prompts

Training Details

Method: QLoRA (4-bit NF4) via Unsloth
LoRA: r=16, alpha=16, attention modules only
Seq length: 2048
Epochs: 2
Hardware: Kaggle T4 GPU

Github

Link : https://github.com/Avijit0001/GemmOpus

Usage

python

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

base = AutoModelForCausalLM.from_pretrained("google/gemma-4-E2B-it", load_in_4bit=True)
model = PeftModel.from_pretrained(base, "AvijitPaul/gemmopus-E2B-reasoning-distill")
tokenizer = AutoTokenizer.from_pretrained("AvijitPaul/gemmopus-E2B-reasoning-distill")