GLM-5.2

As of October 19, 2026, GLM-5.2 by Z.ai (Zhipu AI) costs $1.40 per 1M input tokens and $4.40 per 1M output tokens, with a 1M-token context window. It ranks 8th of 29 models on the TensorFeed Intelligence Index's older 2024 benchmark set, with a score of 75.6 (as of October 22, 2026).

GLM-5.2 is Z.ai's (Zhipu AI) flagship, shipped June 13, 2026. It is a 744 billion parameter mixture-of-experts model with a 1 million token context window, 131K max output, and a Max-effort reasoning mode, released under an MIT license. API pricing is roughly $1.40 per million input tokens and $4.40 per million output, which is on the order of 80 percent below Claude Opus 4.8. Z.ai reports 62.1 on SWE-Bench Pro, which would top GPT-5.5, and the model sits fourth overall and first among open-weight models on the Artificial Analysis Intelligence Index at 51; treat both as vendor-reported. The detail that makes it strategically interesting is the training run: roughly 100,000 Huawei Ascend 910B chips with no Nvidia silicon in the loop, at an estimated $25 million all-in. Reporting puts it at something like 40 percent of developer tokens on OpenRouter. Self-hosting at full precision needs about 1.5TB of GPU memory, roughly nineteen H100s, so the MIT license buys far more freedom than most teams can actually exercise.