CGTN News – Chinese AI models have taken the top five spots in weekly token usage, according to latest data from OpenRouter, a global large-language model (LLM) aggregation leaderboard.
From July 28 to August 3, AI models worldwide processed 56.8 trillion tokens. Chinese models accounted for 28.13 trillion tokens, compared with 4.38 trillion for US models. That’s 14 weeks in a row that China has held the lead.
Nine of the world’s top 10 most-used models last week were Chinese. DeepSeek’s V4-Flash topped rankings with 7.1 trillion tokens in weekly usage, followed by Xiaomi’s MiMo-V2.5 with 5.4 trillion, Tencent’s Hy3 with 4.89 trillion, DeepSeek’s V4-Pro with 3.1 trillion and Z.AI’s GLM-5.2 with 2.87 trillion.
Cost comparison
DeepSeek V4-Flash charges $0.14 per million input tokens and $0.28 per million output tokens. Artificial Analysis pegs its average cost at 3 cents per test, compared with 86 cents for Moonshot’s Kimi K3, $1.86 for OpenAI’s GPT-5.6 Sol and $3.15 for Anthropic’s Claude Fable 5.
The comparison provides a more realistic measure of value than pricing alone because it accounts for the amount of data a model must process and generate to complete a task. A model with low headline price can still prove expensive if it requires significantly more steps to produce an answer.
On the Intelligence Index, DeepSeek V4-Flash scored 50 out of 100, level with Google’s Gemini 3.6 Flash, and one point behind Meta’s Muse Spark 1.1 and Z.AI’s GLM-5.2. Moonshot’s Kimi K3 scored 57 while Anthropic’s Claude Opus 5, Fable 5 and OpenAI GPT-5.6 scored nine or more points higher.
Artificial Analysis uses a nine-benchmark framework spanning coding, reasoning and workplace-style assignments, to compile model scorecards.
What’s given DeepSeek an edge in the global developer market is engineering: better inference, dynamic KV caching, traffic scheduling and long-context optimization.
More overseas companies are integrating DeepSeek APIs, while more international developers are adopting its platform.
Separately on Monday, Alibaba unveiled its largest and most capable AI model to date, the Qwen3.8-Max, which is not far behind in size when compared with an offering from domestic rival Moonshot AI launched last month.
For millions of AI developers worldwide, the shift is already becoming clear: more choices, lower costs and fewer barriers to innovation are reshaping the global AI ecosystem.
(With input from Reuters)
