P
PiazBench

Methodology

How PiazBench evaluates and scores AI coding models across all benchmarks.

Composite Score Formula

Quality = mean of 6 quality benchmarks (UI, Debugging, Security, Refactoring, Reasoning, Viet Bench). Vibe = 70% Quality + 30% Speed (normalized). Speed is normalized to 0-100 based on the highest throughput model.

Quality

mean(UI, Debugging, Security, Refactoring, Reasoning, Viet Bench)

Vibe

0.7 × Quality + 0.3 × Speed_norm

Data Sources

All scores are sourced from publicly available benchmarks and vendor reports. Speed data from Artificial Analysis. Vietnamese Bench scored manually by Piaz team.

Benchmark Details