Sources
Z.ai Launches GLM-5V-Turbo: 744B MoE Vision Coding Model Scores 94.8 on Design2Code vs Claude's 77.3
Zhipu AI's GLM-5V-Turbo is a 744B-parameter MoE (40B active) vision coding model with native multimodal fusion via CogViT encoder, trained on 28.5T tokens. It natively processes images, videos, design drafts, and document layouts in a 200K context window, mastering the perceive-plan-execute loop for autonomous coding. On the Design2Code benchmark (reproducing UI mockups in code), it scores 94.8 vs Claude Opus 4.6's 77.3 — a significant gap in visual-to-code workflows. Available on OpenRouter and integrated with TRAE and Vision Arena.
↳ Follow the thread