Meta hits #4, GLM 5.1 tops coding charts

GLM 5.1 proves that open-source models can now dominate proprietary giants in technical domains like software engineering. Its #3 coding rank puts it ahead of Claude 4.6 Sonnet, marking a massive win for the open-weights community. While the 754B parameter MoE architecture requires significant hardware that makes local deployment challenging for most, its training on Huawei Ascend 910B chips signals a successful decoupling from NVIDIA's hardware monopoly. Meanwhile, Meta's #4 spot confirms their continued dominance in human preference tuning for general chat tasks.