Alibaba Releases and Open-Sources Three Qwen 3.5 Mid-Size Models, Achieving High Performance Without Parameter Stacking

On February 25, Qwen officially released and open-sourced of the latest mid-size models in the Qwen 3.5 series, including Qwen 3.5-35B-A3B, Qwen 3.5-122B-A10B, and Qwen 3.5-27B. Through architectural optimization, improved data quality, and reinforcement learning, these three models no longer rely on simple parameter stacking, achieving performance that surpasses larger-scale previous-generation models. The new Qwen 3.5 models can be deployed directly on consumer-grade GPUs, making them highly developer-friendly. Currently, Qwen 3.5-Flash based on Qwen3.5-35B-A3B has been launched on Alibaba Cloud Bailian, with input costs as low as 0.2 yuan per million tokens.