GLM-5.2: The New Open-Weights AI Model Leading on Intelligence vs. Cost
Z.ai's GLM-5.2 tops the Artificial Analysis Intelligence Index while offering the lowest cost per task at its performance tier, a key consideration for optimizing AI operational expenses.
Practical Summary
A detailed benchmark comparison reveals GLM-5.2 as a leading open-weights model, achieving an Intelligence Index score of 51. It is positioned on the Pareto frontier for intelligence versus cost, with a reported cost of ~$0.46 per task. This data provides a direct, evidence-based metric for businesses evaluating cost-effective AI models for specific workloads.
Why It Matters
For businesses focused on AI cost reduction, GLM-5.2's benchmark performance and pricing allow for a direct comparison against other leading models. The Pareto frontier position suggests it delivers high intelligence at a specific cost point, enabling data-driven decisions about which model to deploy based on required task complexity and budget constraints. The availability across multiple providers also offers supply chain flexibility.
Understanding GLM-5.2's Benchmark Position
The Artificial Analysis Intelligence Index v4.1 scores models on capability. GLM-5.2 leads the open-weights category with a score of 51, surpassing MiniMax-M3 (44) and DeepSeek V4 Pro (44). This benchmark is useful for gauging relative model strength on a standardized set of tasks.
Analyzing the Cost Per Task Metric
The 'Cost per Task' is a critical metric for workflow cost optimization. GLM-5.2's cost is ~$0.46 per task. For context, models with lower benchmark scores like MiniMax-M3 cost ~$0.18 per task, while models with similar scores in the proprietary category may cost significantly more. This metric helps estimate operational cost for batch processing or high-volume API use.
Key API Pricing and Token Usage
Direct API pricing from Z.ai is: $1.4 per million input tokens, $0.26 per million cache hit tokens, and $4.4 per million output tokens. Note that GLM-5.2 uses an average of 43,000 output tokens per task (up from 26,000 for GLM-5.1), which directly impacts the output token cost component of your total expense per task.