K3 drives Moonshot global pivot

Moonshot’s center of gravity moved from a China consumer app into a global model business. The domestic user lead that Kimi built with long context proved fragile once DeepSeek won Chinese app attention, but K3 gave Moonshot a new route to growth through overseas subscriptions, APIs, and hosted deployments, where buyers care less about app rankings and more about model capability, deployment options, and workflow fit.1 2 3

• The reversal makes the timeline clear. Kimi’s monthly active users fell from more than 20 million to roughly 9 million by the end of 2025 as DeepSeek gained users, then overseas revenue passed domestic revenue in February 2026. Revenue growth came after Moonshot cut ads and leaned into business distribution instead of trying to win a consumer traffic war in China.1 2
• K3 mattered because it traveled well outside Moonshot’s own app. The model was released as open weights, sold through Moonshot’s API, and hosted by providers like Fireworks AI and Together AI. That let overseas developers and enterprises use Kimi through familiar infrastructure, keep data handling with their chosen host, and build coding or legal agents without adopting a Chinese consumer product.1 2 3
• This is also a competitive shift from chatbots to production workloads. DeepSeek still pressures Moonshot on low cost reasoning inside China, but internationally Moonshot competes more with OpenAI and Anthropic for coding and knowledge work budgets. The real battle is over which model sits inside a developer tool, research workflow, or enterprise agent, because that is where recurring token spend compounds.1 2 3

The next step is turning this overseas demand spike into durable distribution. If Moonshot keeps placing Kimi inside cloud marketplaces, inference providers, and work products like Kimi Code and Kimi Work, international revenue can keep widening its lead over domestic revenue and push the company further toward an enterprise infrastructure role rather than a consumer app company.1 3