Alibaba Releases Qwen4-Max, with the Preview Available First on Multiple Products
On November 16, the Qwen team officially released its flagship model, Qwen4-Max, saying it is based on the new Qwen4 architecture and uses a sparse MoE with 2.4 trillion parameters, activating 71B per token. It features native vision capabilities and a 1-million-token context window. Earlier, on November 11, the Qwen4-Max preview launched on Alibaba Token Plan (with a limited-time offer as low as 10% of the regular price) and became available on products including the Qwen App, Qwen Office, Qoder, and QwenStudio. The Qwen Go! membership in the Qwen App has been offered at a limited-time discounted price starting at 9.9 yuan since November 11. Following the official launch, the Qwen App said members could try the new model for free during its launch period, with Qwen4-Max usage not counting against membership quotas. See the product page for the promotion period. The Qwen team says Qwen4-Max is more efficient in token usage and cost than Qwen3.8-Max: internal evaluations show an average cost per task reduction of more than 60%. In the Work360 evaluation, token usage fell by about 24% (Xhigh) and 35% (High), while average cost per task dropped from $4.43 to $1.67. Its Work360 benchmark score is 52.69, higher than …
Latest developmentsQwen Releases Its Next-Generation Flagship Model, Qwen4-Max
