GLM-5.2 tops frontend coding benchmarks

Z.ai released GLM-5.2, an MIT-licensed open-weight model with a 1M-token context window and two reasoning-effort modes. Third-party evals rank it just behind Opus 4.8 as the best coding model worldwide, and it beats all Opus versions at frontend coding. The 744B parameter model uses improved DeepSeek Sparse Attention for ultra-long contexts.