On February 16, Alibaba took the wraps off its latest open-source marvel: the QianWen Qwen3.5-Plus, a large-scale model whose capabilities stand toe-to-toe with those of Gemini 3 Pro, cementing its position as the most potent open-source model globally. QianWen 3.5 has undergone a thorough enhancement in its foundational model architecture. The Qwen3.5-Plus variant is equipped with a staggering 397 billion parameters, yet it cleverly activates only 17 billion, demonstrating remarkable efficiency by achieving superior results compared to the Qwen3-Max model, which boasts a trillion-level parameter count. Simultaneously, it slashes deployment memory usage by 60% and significantly boosts inference efficiency, with the peak inference throughput soaring up to 19 times higher than the original.

Furthermore, the API pricing for Qwen3.5-Plus is set at a mere 0.8 yuan per million tokens, representing just 1/18th of the cost associated with Gemini 3 Pro.