kimi-k2-thinking-turbo

Kimi-k2-Thinking-Turbo is a high-performance large language model launched by **Moonshot AI**, part of the **Kimi K2 series**. This series of models features a **Mixture-of-Experts (MoE) architecture**, boasting approximately **1 trillion total parameters (1T)** and **32 billion active parameters (32B)**. While maintaining robust reasoning capabilities, it achieves efficient computation and inference performance, with output speeds boosted to 60–100 tokens per second.