An enhanced, lightweight code-reasoning model built on top of LiquidAI's LFM-2.5-2.6B. Fine-tuned to boost step-by-step reasoning and coding capabilities without sacrificing overall model performance.

📌 Quick Summary

Feature	Details
Developed by	Schnuckade
Base Model	LiquidAI/LFM2.5-2.6B
Fine-Tuning Framework	Unsloth 🦥
Dataset Used	`Banaxi-Tech/Deepseek-V4-Reasoning-Code-2500`
License	Apache-2.0
Language	English

💡 About the Model

LFM-2.5-Coder-2.6B is a targeted, lightweight fine-tune designed to make the original Liquid AI base model a bit smarter when dealing with code generation and problem-solving.

🧠 Reasoning Focus: Helps the model "think" through coding problems step-by-step.
⚡ Lightweight & Efficient: Trained specifically to boost coding skills while retaining general utility.
📦 Multi-Format: Includes full FP16/BF16 safetensors alongside quantized GGUF weights (Q4_K_M) for easy local deployment.

🛠️ Usage & Integration

Transformers / Unsloth

python

from unsloth import FastLanguageModel

model, tokenizer = FastLanguageModel.from_pretrained(
    model_name = "Schnuckade/LFM-2.5-Coder-2.6B",
    max_seq_length = 2048,
    load_in_4bit = True,
)