Kimi K3 vs DeepSeek V4 Pro vs GLM-5.2: Open Trillion-Scale MoE Models Compared to Benchmarks, Licensing, and Supply Costs

Three Chinese labs now hold the top of the open weight leaderboard. Moonshot AI’s Kimi K3, DeepSeek V4 Pro, and Zhipu AI’s GLM-5.2 are all outstanding Mixture-of-Experts (MoE) models with context windows of one million tokens. Each targets long-horizon coding and agent workloads. This article compares them on three axes that the AI ​​team decides on: rated power, license terms, and cost of use.

‘Trillion-parameter’ equals Kimi K3 (2.8T) and DeepSeek V4 Pro (1.6T). GLM-5.2 is a total of 744B, so it is the smallest of the three in total parameters. It gets its place because it was leading the open weight field before the K3 was introduced.