A recent study reveals that Chinese AI models account for 89% of global enterprise spending on open-source models. Moonshot AI's Kimi leads the pack with a 61% share, followed by DeepSeek at 15% and Zhipu GLM at 11%. Among Western models, the Meta Llama series holds the highest share but captures only 7%. Currently, open-weight models handle just 19% of enterprise token usage, while closed-source models continue to dominate the enterprise AI market—Anthropic and OpenAI together command an 86% share of closed-source model spending.
