Alibaba's Qwen team launched Qwen4-Max, built on a new Qwen4 architecture with 2.4 trillion backbone parameters, 71B activated per token, plus 328B N-gram embedding parameters, native vision and a one-million-token context window. The company claims broad gains over Qwen3.8-Max on coding, office and long-horizon tasks while cutting token consumption by roughly 24%-49% and slashing API pricing, with average Work360 task cost down 58%.
Termsslug: n-gram-embedding
N-gram Embedding
A structural design adding contiguous n-gram segment embeddings as parameters separate from the backbone.
Curated Intelligence
Qwen 官方博客 (通义)T1·Models
88 pts
千问AI平台 (微信公众号)T2·Models
78 pts
Alibaba's Qwen team has officially launched Qwen3.8-Flash, a multimodal MoE model featuring the new 'Next' architecture with 100B total parameters but only 6B active. By leveraging QSA sparse attention and Gated Residual techniques, it reduces computational overhead significantly, achieving over 8x speedup in long-context scenarios while offering ultra-low pricing at $0.11 per million input tokens.