Base Model Blind Spots Analysis: Nanbeige4-3B-Base
Model Tested
Model: Nanbeige4-3B-Base (Strictly the Base pre-trained model, avoiding the SFT/RL aligned Nanbeige4.1-3B iteration).
Release Date: Early 2026
Modality: Text
Parameters: 3B
Loading the Model
The model was loaded using Hugging Face transformers on a Google Colab T4 GPU instance with bfloat16 precision to fit into 16GB of VRAM:
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM… See the full description on the dataset page: https://huggingface.co/datasets/KhalifaBouneb/Nanbeige4-3B-Blindspots.
