GLM SEA

Z.ai's GLM waters, charted since 2019 from Beijing. GLM-5.3 advances software-engineering and agent capabilities; GLM-4.6 (late September 2025) marked the first FP8 and Int4 quantization on Cambricon chips and supports native FP8 on Moore Threads GPUs; GLM-4.5 (July 2025) runs on eight NVIDIA H20 chips. Open-weight under MIT since July 2025.