Top-rank quality at 1/28 the cost per successful task
↑ Navigator Index (0–100) — higher is better
Value sweet spotClears the bar for less
FrontierClears the bar, at a price
OutclassedCheap, but under the bar
Paying for the badgeCostly and under the bar
Quality bar · Index 70
Human baseline · 88.0 · unranked reference
Model M · low94.2 · $0.0042
Model D · high95.5 · $0.12
Model D · low94.4 · $0.045
Model L · low54.8 · $0.0034
020406080100
$0.0020$0.0050$0.010$0.020$0.050$0.10$0.20
← Cost per successful task (log scale) — lower is better
Navigator Bench · Index v2.0 · data through 14 Oct 2026Line: Pareto frontier · quadrants split at Index 70 and the median $0.013Demo data
6 models share rank 1 (none is significantly better than another on the same tasks); Model M · low effort (Index 94.2) costs $0.0042 per successful task, Model D · high effort (Index 95.5) $0.12.
1/5 · Quality vs cost — ← → scenes · space play/pause · R replay · F full screen · Esc leaves full screen, then the presenter