@Kimi_Moonshot hit ~$1.2B ARR in September (Nomura estimate), up from $1B in August. Targeting $2B by year-end.
The real moneymaker? Not the flagship—it's the cheaper K2.8 Preview:
• K2.8 Preview (launched 9/11) drives 40-45% of revenue vs K3's 30-35%
• Lower price point, optimized for coding and daily agent tasks
• Gross margins: K3 at 35-40%, budget models at 20-25%
• API is the primary revenue stream, with over half coming from overseas
Classic case of volume over margin—cheaper inference winning the adoption game while flagship handles premium workloads. The split suggests developers care more about cost-per-task than raw capability for most use cases.
The real moneymaker? Not the flagship—it's the cheaper K2.8 Preview:
• K2.8 Preview (launched 9/11) drives 40-45% of revenue vs K3's 30-35%
• Lower price point, optimized for coding and daily agent tasks
• Gross margins: K3 at 35-40%, budget models at 20-25%
• API is the primary revenue stream, with over half coming from overseas
Classic case of volume over margin—cheaper inference winning the adoption game while flagship handles premium workloads. The split suggests developers care more about cost-per-task than raw capability for most use cases.