--- base_model: - saishshinde15/TBH.AI_Base_Reasoning tags: - text-generation - transformers - causal-lm - reasoning - sft - gguf license: apache-2.0 language: - en --- # **TBH.AI Vortex Reasoning (GGUF)** - **Model Name:** `saishshinde15/TBH.AI_Vortex_Reasoning_GGUF` - **Developed by:** TBH.AI - **License:** Apache 2.0 - **Fine-tuned from:** [TBH.AI_Base_Reasoning](https://huggingface.co/saishshinde15/TBH.AI_Base_Reasoning) - **Available in:** **16-bit and 4-bit GGUF formats** --- ## **Overview** TethysAI Vortex Reasoning is an **experimental model** designed to replicate the **advanced reasoning** abilities of **TBH.AI_Base_Reasoning**, which was originally enhanced using **GRPO**. Instead of GRPO, this model was fine-tuned with **high-quality structured data** using **high-end Supervised Fine-Tuning (SFT)** to replicate the step-by-step **thinking and self-questioning** mechanisms seen in models like **DeepSeek-R1**. This model has been optimized for **efficient inference in GGUF format**, allowing for **deployment on CPU-based systems** and **lightweight edge devices** without sacrificing reasoning capabilities. --- ## **Why This Model Stands Out** 🔹 **Advanced Self-Reasoning:** - The model **questions itself** internally before arriving at an answer. - Similar to **DeepSeek-R1**, it follows a **structured reasoning process**. - Uses **** and **** tokens internally, though they may not always be explicitly visible in responses. 🔹 **No GRPO, Only High-End SFT:** - Instead of **GRPO**, the model **learns structured reasoning directly** from fine-tuned data. - **Demonstrates logical breakdowns, multi-step problem-solving, and contextual understanding.** - Achieves results **comparable** to the base model without reinforcement learning. 🔹 **Optimized for GGUF Inference:** - Available in **both 16-bit and 4-bit GGUF**, enabling **fast and memory-efficient execution** on **CPUs**. - Ideal for **on-device deployment**, including **edge computing, embedded AI, and AI assistants.** --- ## **Usage** # Use the below prompt for the best results: ```python You are an advanced AI assistant. Provide answers in a clear, step-by-step manner. ```