A November 8 Menlo Ventures report cited says Chinese AI models take 89 percent of global enterprise open-source model spending
The Menlo Ventures report shows Kimi at 61 percent of global enterprise open-source model spending, followed by DeepSeek at 15 percent and GLM at 11 percent, while Meta Llama, the top Western open model, holds just 7 percent.
On November 8, citing a South China Morning Post report from two days earlier, it was reported that a recent study by Silicon Valley venture firm Menlo Ventures found Chinese AI models led by Moonshot AI's Kimi family account for 89 percent of global enterprise spending on open-source models, close to nine tenths. Kimi led with 61 percent of enterprise open-source model spending, followed by DeepSeek at 15 percent and Zhipu's GLM at 11 percent.
By comparison, Meta Platforms' Llama family is the highest-share Western model in this category, but it accounts for only 7 percent of enterprise open-source model spending. The firm also noted that open-weight models handle 19 percent of enterprise token usage, while closed models still dominate the enterprise AI market, with Anthropic and OpenAI together taking 86 percent of spending.
The metric is the share of enterprise spending on open-source models and related products and services, not model downloads, developer counts or benchmark rankings. The fact that enterprises are willing to pay usually means a model has entered production or near-production systems. Another detail from the report says enterprise spending on AI applications will grow from 33 billion dollars a year earlier to about 177 billion dollars in 2026.