Alibaba's Qwen3-Coder-Next achieves 70.6% on SWE-bench with efficient MoE architecture

The Qwen3-Coder-Next model, an 80 billion parameter Mixture-of-Experts model from Alibaba's Qwen team, has demonstrated impressive efficiency by achieving 70.6% on the SWE-bench Verified benchmark with only approximately 3 billion active parameters per inference pass. This allows it to offer performance comparable to frontier coding agents while requiring hardware resources similar to a 7 billion parameter model. The model supports a 256K context window, making it suitable for complex coding tasks, and can be set up locally using Ollama for an OpenAI-compatible API. AI

IMPACT Sets a new bar for efficient coding models, potentially lowering hardware barriers for advanced AI-assisted development.

RANK_REASON New model release with benchmark performance from a major lab. [lever_c_demoted from frontier_release: ic=1 ai=1.0]

Read on dev.to — LLM tag →

• Alibaba
• Apache 2.0
• Continue.dev
• GGUF
• GPT-4o
• Ollama
• Qwen3-Coder-Next
• Qwen team
• RunPod
• SWE-bench Verified

• model release
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →