Benchmark leaderboard
Nine typed decision models, benchmarked on the public JevBench tiers.
Highlights
Results by tier
Click a column heading to sort. Ranked by mean accuracy across the three tiers.
What fits in the GPU
Measured memory per model on this RTX 5090, and how many can run at once. Pick models to see the total, or use a preset.
Accuracy against speed
Hard-tier accuracy against median latency. Up and to the left is better.
Accuracy by decision family
Where each model is strong or weak. Each family holds only a handful of decisions per tier, so treat single cells as indicative.