Qwen3-30B-A3B
🧠 AI ModelQwen
30B-parameter MoE model with only 3B activated per token, open-source under Apache 2.0.
Qwen3-30B-A3B is a state-of-the-art text-generation model using a Mixture-of-Experts (MoE) architecture developed by Qwen. It has 30 billion total parameters, of which only 3 billion are activated per forward pass, enabling high performance with lower computational cost. The model is fine-tuned from the base version (Qwen3-30B-A3B-Base) and released under the permissive Apache 2.0 license. It achieves competitive results on various benchmarks while being efficient for deployment. The architecture employs top-k routing, specialized expert modules, SwiGLU activations, and rotary position embeddings. It is part of the Qwen3 family, which includes models from 0.5B to 236B parameters. Available on Hugging Face with over 1.5 million downloads and 900 likes, it is designed for long-context understanding and a wide range of natural language tasks.
💡Highlights
- ├─30B MoE, 3B active per token
- ├─Apache 2.0 license
- └─1.5M downloads, 900 likes
🎯For
- ├─AI researchers
- ├─Machine learning engineers
- └─Open-source enthusiasts