DeepSeek V4 and Alibaba Qwen 3.8 Launch Boost Chinese AI Race

Chinese AI firms have unveiled two major large‑language‑model releases. DeepSeek announced the full‑version V4, touting a 40‑50× lower computation cost than OpenAI or Anthropic and a pay‑as‑you‑go pricing model that undercuts rivals by up to 30‑70 times. The model is positioned as a cost‑effective alternative to US‑based closed‑source systems and is claimed to approach the performance of top‑tier models such as Fable 5.

Alibaba released Qwen 3.8 max as an open‑source offering, emphasizing a “conditional open‑source” strategy after internal debate over the economics of fully open flagship models. The release comes amid intense competition with the newly disclosed Kimi K3, a 2.8‑trillion‑parameter model that, while cheaper than Western alternatives, still faces high operating costs. Alibaba’s Qwen aims to leverage its cloud infrastructure to provide a lower‑price, faster model while retaining a share of the ecosystem through its cloud services.

Both launches highlight a shifting Chinese AI landscape where cost, openness, and cloud integration are key battlegrounds against U.S. incumbents.