Alibaba Open Sources Qwen3.6-35B-A3B Rivaling Frontier Models With 3B Active Parameters

Apr 16, 2026 · Updated Apr 30, 2026

Alibaba open-sourced Qwen3.6-35B-A3B, a sparse Mixture-of-Experts (an architecture activating only a fraction of parameters per task) model. It features 35 billion total parameters but uses only 3 billion active parameters during inference. This release includes native multimodality and a hybrid reasoning mode that supports extended thinking.

This update shifts the efficiency frontier by delivering agentic coding (AI that autonomously plans and executes multi-step tasks) performance on par with models ten times its active size. It bridges the gap between lightweight models and frontier reasoning systems, matching Claude Sonnet 4.5 on vision benchmarks and spatial intelligence.

You can integrate the model into agentic workflows using tools like OpenClaw, Claude Code, or Qwen Code. Weights are available on Hugging Face and ModelScope under an Apache 2.0 license for local deployment. It is also accessible via the Alibaba Cloud Model Studio API as qwen3.6-flash.

View the full update on qwen.ai

Qwen

@Alibaba_QwenApr 16

⚡ Meet Qwen3.6-35B-A3B：Now Open-Source！🚀🚀 A sparse MoE model, 35B total params, 3B active. Apache 2.0 license. 🔥 Agentic coding on par with models 10x its active size 📷 Strong multimodal perception and reasoning ability 🧠 Multimodal thinking + non-thinking modes Efficient. Powerful. Versatile. Try it now👇 Blog：https://t.co/EXx5y466su Qwen Studio：https://t.co/bg4tAU1p74 HuggingFace：https://t.co/w4pDX14DZS ModelScope：https://t.co/SuRyLzdQiO API（‘Qwen3.6-Flash’ on Model Studio）：Coming soon～ Stay tuned

1.2k8k

View on X