Task intelligence (Artificial Analysis) vs operational/logistic quality (OpenRouter, 1-wk) vs cost.Hover any point for details.
Dashed lines = medians. Top-right = strong on both. Color = 7:1 blended cost.
Quality = min(percentile-intel, percentile-logistic). Up-left = best value. Line = value frontier.
| Model | Intel | Logistic | Quality | Cost $/M |
|---|
Data: OpenRouter performance stats (throughput, E2E latency, tool-call & structured-output error, cache hit, uptime - volume-weighted across providers, 1-week) and Artificial Analysis benchmarks (Omniscience, Briefcase, GDPval, LCR, tau2-Telecom). Intelligence & logistic scores are min-max normalized to /10; quality is the min of the two percentile ranks. Cost is 7:1 input:output blended $/M. 109 models with all three dimensions shown.