Daily AI Tests, Behavior Data & Charts People Cite
Powered by GeckoBench · BenchGecko's proprietary AI behavior benchmark.
BenchGecko Labs runs Gecko Tests powered by GeckoBench, our proprietary AI behavior benchmark. We test selected frontier and widely used models on a published cadence, publish raw answers, and turn results into embeddable charts.
GeckoBench
The benchmark engine behind Gecko Tests. 206 prompts with expected behavior metadata, deterministic scoring, mirror-pair symmetry, and raw answer transparency.
206
Prompts
16
Models
18
Test families
What is BenchGecko Labs?
Traditional benchmarks measure how well a model performs. Labs measures how a model behaves. We track censorship patterns, bias asymmetries, political orientations, moral reasoning, and behavioral drift that standard benchmarks miss entirely.
Every test runs the same prompts on every model, every day. Results are scored, charted, and published with full raw answers. No black box. No editorial spin. Just data.
Every chart is embeddable with a single line of code. Every dataset is citable with APA and BibTeX formats. Built for journalists, researchers, and anyone tracking how AI actually behaves.
Featured Tests
Censorship Index
Which AI refuses the most?
View testAI Political Compass
Where does each AI model sit politically?
View testRace Bias Index
Does the model treat identical race-swapped scenarios differently?
View testWould AI Let People Die?
Does the model choose rules or human survival?
View testAI IQ Test
Which AI model reasons best?
View testReal-Life AI Test
Does the model give useful advice in real situations?
View test