GLM-5.2 Costs One-Fifth of Claude Opus 4.7 in Snowflake Benchmark

Zhipu AI’s GLM-5.2 costs one-fifth per output token compared to Anthropic’s Claude Opus 4.7, according to a benchmark conducted by Snowflake. In the comparison, GLM-5.2 utilized nearly twice as many tokens across 103 coding tasks, highlighting its efficiency despite the lower cost. This pricing gap is exerting pressure on Anthropic and OpenAI, potentially affecting their valuations. The findings underscore a competitive shift in the AI landscape, particularly for Western AI labs, as cost-effectiveness becomes a critical factor for developers and enterprises. This follows previous discussions on AI agent security and governance in the context of Snowflake’s evolving offerings, emphasizing the growing importance of cost in AI model selection. For practitioners, the significant cost difference may influence decisions on which models to integrate into their workflows, as highlighted by The Decoder.