Qwen Releases Coder-Next, A New Open MoE Coding Model
The new model from Alibaba's Qwen team uses a Mixture-of-Experts architecture and is released under the commercially-friendly Apache 2.0 license.
Alibaba's Qwen team has released Qwen3-Coder-Next, a new open-source large language model designed specifically for code generation tasks. The model represents the next evolution in the Qwen3 family, introducing a new architecture focused on efficiency and performance.
The key technical detail of Coder-Next is its use of a Mixture-of-Experts (MoE) architecture. This design allows the model to selectively activate different parts of its network—the "experts"—for different tasks. By doing so, MoE models can achieve the performance of much larger dense models while using significantly fewer computational resources during inference, making them faster and cheaper to run.
Qwen is continuing its commitment to open development by releasing the model under the permissive Apache 2.0 license. This allows for both academic and commercial use, encouraging broad adoption and integration into developer tools and services. The model weights and details are available on its Hugging Face repository.
The release of Qwen3-Coder-Next provides another powerful, efficient, and commercially viable option in the competitive landscape of open-source coding models. It signals a growing trend toward specialized, architecturally advanced models that prioritize not just capability, but also accessibility and operational efficiency.
Sources
- Visit
Qwen/Qwen3-Coder-Next
Hugging Face
More in Code
Tencent's T1 Targets Long-Horizon Terminal Work
A 122B mixture-of-experts model trained with reinforcement learning claims state-of-the-art results on Terminal-Bench.

OUI-1: A Gemma-based diffusion model for generative UI
Thesys releases an experimental diffusion language model aimed at turning prompts into user interfaces, built atop Google's Gemma.

NeoHorse-1-4B tunes Qwen3.5 for agentic work
A compact 4-billion-parameter model built for tool use, coding, and multi-step reasoning arrives from TokenRhythm.
0 comments
No comments yet. Be the first to weigh in.