Qwen 4 Max drops with multimodal reasoning
Alibaba's Qwen 4 Max flagship multimodal model is now live, featuring a 2.4 trillion-parameter MoE architecture and a 1M-token context window. It promises significantly higher efficiency and lower costs compared to its predecessor.
Qwen 4 Max marks Alibaba's continued push into flagship reasoning models, although the shift to a hosted-only model is a notable departure from their open-weights legacy.
- –2.4T total parameters with 71B active per token MoE architecture
- –Massive 1M-token native context window for complex, multi-step reasoning
- –~60% reduction in task execution costs vs. Qwen 3.8-Max
- –Closed-weights approach contrasts with previous open-source Qwen variants
DISCOVERED
9h ago
2026-11-17
PUBLISHED
9h ago
2026-11-17
RELEVANCE
AUTHOR
Av1dlive