Alibaba releases Qwen3.6-35B-A3B model demonstrating competitive performance against Google Gemini at fraction of cost

Alibaba released Qwen3.6-35B-A3B on April 16, 2026, featuring 35 billion total parameters with 3 billion active during inference. The model demonstrates competitive agentic coding performance against larger models like Google's Gemini 3.1 Pro while offering significant cost advantages at $0.38 per million input tokens compared to Gemini's $2.00. It also shows strong multimodal perception and reasoning abilities.