Nvidia Outperforms AMD in AI Coding Agent Benchmark on Cost Efficiency

A recent open benchmark by SemiAnalysis demonstrated that Nvidia's GLM 5.3 configuration runs five times cheaper than AMD's for real coding-agent traffic, with the cost difference attributed to serving software.