Kimi K2.6 Outperforms Zhipu AI's GLM-5.1 in Coding Benchmarks
Cheaper per token and slightly faster on code tasks.
Moonshot AI's latest model, Kimi K2.6, released on April 20, 2026, has achieved slightly better performance on coding benchmarks compared to Z.ai's (formerly Zhipu AI) GLM-5.1, which launched on April 7, 2026. While Kimi K2.6 edges ahead in coding tasks, it also comes with a lower cost per token for serving, making it a more economical choice for developers and enterprises. The exact benchmark scores were not disclosed, but the margin is reportedly narrow, suggesting intense competition in the Chinese AI landscape.
GLM-5.1 is a massive 754B-parameter model that was trained exclusively on 100,000 Huawei Ascend 910B chips, avoiding NVIDIA GPUs entirely. This underscores China's strategic push to reduce reliance on Western semiconductor technology and build an independent AI infrastructure. Meanwhile, Kimi K2.6's cheaper serving costs could make it more attractive for real-world applications, especially in cost-sensitive environments. The rivalry between Moonshot AI and Z.ai highlights the rapid pace of innovation in China's AI sector, with both models pushing boundaries in performance and efficiency.
- Kimi K2.6 outperforms GLM-5.1 slightly on coding benchmarks
- Kimi K2.6 has lower cost per token for serving than GLM-5.1
- GLM-5.1 is a 754B-parameter model trained on 100,000 Huawei Ascend 910B chips without NVIDIA GPUs
Why It Matters
China's AI models are advancing rapidly, with domestic hardware and cost efficiency becoming key differentiators.