Qwen3-0.6B

Introduction

Qwen3 is the latest large language model in the Qwen series, offering both dense and Mixture-of-Experts (MoE) architectures. It features seamless switching between thinking and non-thinking modes, enhanced reasoning and instruction-following abilities, superior human preference alignment for engaging conversations, strong agent capabilities for tool integration, and support for over 100 languages with excellent multilingual performance.

  • Type: Causal Language Model

  • Training Stage: Pretraining & Post-training

  • Number of Parameters: 0.6B (0.44B non-embedding)

  • Number of Layers: 28

  • Number of Attention Heads (GQA): 16 for Q, 8 for KV

  • Context Length: 32,768

Available NPU Models

Base Model

qwen3-0.6B-ax630c

The base model providing a 128 context window and a maximum output of 1,024 tokens.

Support Platforms: LLM630 Compute Kit, Module LLM, and Module LLM Kit

  • 128 context window

  • 1,024 max output tokens

  • ttft 361.81ms

  • avg-token/s 10.28

Install
apt install llm-model-qwen3-0.6b-ax630c

Manual installation: Click here to download llm-model-qwen3-0.6b-ax630c