GLM-5.2 is an open-source flagship model supporting a 1M-token context for long-horizon tasks, with advanced coding capabilities, flexible effort levels, and an improved architecture that reduces per-token computational costs by 2.9x compared to its predecessor.