AlibabaのQwen3.8-Max vs DeepSeekのV4-Flash:中国AI価格戦争はゼロに向かっているのか?

alibaba escalated the chinese ai arms race by releasing qwen3.8-max, its largest model yet, utilizing a mixture-of-experts architecture to manage 2.4 trillion parameters while activating only 95 billion at a time. simultaneously, deepseek drew focus with its v4-flash model, specifically targeting market share through inference pricing that undercuts major global competitors. these moves suggest a strategic pivot for chinese labs, prioritizing cost efficiency and infrastructure optimization to scale deployments amid high compute costs. the development pressures global leaders like openai and anthropic, whose api costs remain significantly higher, to re-evaluate their own pricing models.