GLM-5.2 (max), an open weights model released in June 2026, achieves a score of 51 on the Artificial Analysis Intelligence Index, placing it significantly above the 24-point average of comparable models. The model, which features a 753B total parameter count with 40B active parameters, is notable for its speed, reaching 111 output tokens per second, which surpasses the class average of 59. It supports text-to-text modality and offers a 1m token context window (approximately 1500 A4 pages).

Despite its performance capabilities, the model is described as expensive compared to other open weights models in its size category. API pricing is set at $1.40 per 1 million input tokens and $4.40 per 1 million output tokens, while the cache hit price is $0.26 per 1 million tokens, reflecting an 81% discount. The total cost to evaluate the model on the intelligence index reached $867.88.

In comparative benchmarks, GLM-5.2 (max) generated 140M output tokens, a figure indicating it is somewhat verbose when compared to the 110M average. It ranks 15th out of 92 models for speed and 77th for cost-efficiency. The model demonstrates reasoning capabilities and is designed for high-intelligence tasks, though it maintains a higher price-per-task ratio than several competing models currently available on the market.

GLM-5.2 (max), an open weights model released in June 2026, achieves a score of 51 on the Artificial Analysis Intelligence Index, placing it significantly above the 24-point average of comparable models. The model, which features a 753B total parameter count with 40B active parameters, is notable for its speed, reaching 111 output tokens per second, which surpasses the class average of 59. It supports text-to-text modality and offers a 1m token context window (approximately 1500 A4 pages).

Despite its performance capabilities, the model is described as expensive compared to other open weights models in its size category. API pricing is set at $1.40 per 1 million input tokens and $4.40 per 1 million output tokens, while the cache hit price is $0.26 per 1 million tokens, reflecting an 81% discount. The total cost to evaluate the model on the intelligence index reached $867.88.

In comparative benchmarks, GLM-5.2 (max) generated 140M output tokens, a figure indicating it is somewhat verbose when compared to the 110M average. It ranks 15th out of 92 models for speed and 77th for cost-efficiency. The model demonstrates reasoning capabilities and is designed for high-intelligence tasks, though it maintains a higher price-per-task ratio than several competing models currently available on the market.