Chinese AI Models Lag Behind US Rivals by Nine Months

Independent benchmarking firm Artificial Analysis reports that leading Chinese AI models consistently lag behind US rivals by approximately three to nine months in intelligence. Despite the recent release of DeepSeek's v4 Flash and Moonshot AI's Kimi k3, the firm notes that the intelligence gap between the US and China frontiers has remained stable over the last few years.

Alibaba Group announced the release of Qwen 3.8 Max, a model featuring 2.4 trillion parameters. Alibaba claims this new model achieves parity with performance levels seen at Anthropic. The benchmarking analysis observes a strategic difference in distribution, noting that Chinese labs tend to release open-weights models, whereas US labs prioritize proprietary models served via API.

The report also highlights that increasing model intelligence enhances capabilities for both cyber offense and defense. These technical advancements have led to increased interest from the Federal government of the United States, which has implemented export controls in response to these capabilities.