1 in 3 AI models used globally now comes from China

Chinese open-source AI models now account for roughly one-third of global AI usage, according to a 2025 study by OpenRouter, an AI model marketplace, a market position that would have been unthinkable just two years ago. DeepSeek leads that share, with Alibaba's Qwen family close behind. The numbers tell a story Silicon Valley no longer controls.

DeepSeek's newly released V4 model outperforms every other open-source system in generating computer code, according to independent testing by Vals AI, which benchmarks AI performance across the industry. That puts it ahead of fellow Chinese startup Moonshot AI's recently released Kimi 2.6, though Vals AI CEO Rayan Krishnan notes the two are "basically neck-and-neck" across broader tasks.

The gap that matters most is still the one between China and the US. V4 trails Anthropic and OpenAI's top proprietary coding models, but the distance is shrinking with each release cycle. That trajectory is exactly

The timing of DeepSeek's V4 launch is not incidental. In the lead-up to the release, OpenAI and Anthropic publicly accused DeepSeek of using distillation, a technique where engineers query a rival model millions of times to replicate its behaviour without building from scratch. Both companies framed this as unfair appropriation of their proprietary technology.

DeepSeek has not formally responded to those accusations. But the allegations reflect a broader anxiety: that open-source Chinese models are advancing not just through independent research but by learning directly from the closed systems they are racing to beat.

China's embrace of open-source AI is not accidental generosity; it is deliberate strategy. A 2025 study by a US congressional advisory body found that freely available Chinese models have spread rapidly across domestic industries, including robotics, logistics, and manufacturing, generating real-world data that feeds back into improving the systems further.

Globally, developers from Lagos to Kuala Lumpur are adopting Chinese models because they are cheaper to run. Malaysia's deputy minister of communications stated last year that the country's sovereign AI infrastructure would be built on DeepSeek's technology, a concrete sign of the geopolitical foothold open-source is creating.

Three successive US administrations have imposed export controls on advanced chips to limit Chinese AI development. Despite that, Alibaba's Qwen model family has crossed one billion downloads. ByteDance invested $11 billion in AI infrastructure in 2024 alone. The chip restrictions have slowed certain capabilities but have not broken the momentum.

Kevin Xu, founder of Interconnected Capital, a hedge fund focused on AI investment, puts it plainly: "Open source is the soft power of technology of the future."