AaryanK/Youtu-LLM-2B-GGUF

Youtu-LLM-2B-GGUF

<div align="center"> <img src="https://huggingface.co/tencent/Youtu-LLM-2B/resolve/main/assets/youtu-llm-logo.png" alt="Youtu-LLM Logo" height="100px"> </div>

Description

This repository contains GGUF format model files for Tencent's Youtu-LLM-2B.

Youtu-LLM-2B is a highly efficient 1.96B parameter model featuring a Dense MLA architecture and a native 128K context window. Despite its small size, it supports Agentic capabilities and "Reasoning Mode" (Chain of Thought), outperforming many larger models in STEM, coding, and agentic benchmarks.

Evaluation Results

<img src="https://huggingface.co/tencent/Youtu-LLM-2B/resolve/main/assets/generalagenticinstruct.png" alt="Comparison between Youtu-LLM-2B and baselines">

Files & Quantization

To see the available files, please verify the Files and versions tab.

How to Run (llama.cpp)

Note: This model uses the Dense MLA architecture. Please ensure you are using the latest version of llama.cpp to support this architecture correctly.

Recommended Parameters: This model supports two modes. Adjust your temperature accordingly:

• —Reasoning Mode (CoT): Temperature 1.0 (Recommended for complex logic/math).
• —Normal Mode: Temperature 0.7 (Recommended for chat/stability).
• —Context: -c (Supports up to 131072).

CLI Example

./llama-cli -m Youtu-LLM-2B.Q4_K_M.gguf \
  -c 8192 \
  --temp 1.0 \
  --top-p 0.95 \
  -p "User: Explain the theory of relativity.\nAssistant:" \
  -cnv

Server Example

./llama-server -m Youtu-LLM-2B.Q4_K_M.gguf \
  --port 8080 \
  --host 0.0.0.0 \
  -c 16384 \
  -ngl 99