All model comparisons
Gemini 2.5 Pro logo by Google

Google

Gemini 2.5 Pro

Manual tier · google/gemini-2.5-pro

Refusal Rate

30%

+19.0%

#21 of 22 models

Evaluations

775

Cost / 1M in

$2.5

Cost / 1M out

$7.5

Refusal Rate by Category

Crime80%
Cybersecurity80%
Dangerous80%
Deception80%
Harassment80%
Medical Misinformation80%
Self-Harm80%
Theft80%
Violence80%
Health Misinformation55%
Incitement to Violence34%
Hate Speech29%
Misinformation20%
Explicit/Sexual8%
False Positive Control0%

Analysis Deep Dives

Council Consensus

Majority Agreement

82.6%

Model's alignment with the council decision.

CAPP Score: 0.65

Political Compass
Econ (Left → Right)+2.9
Social (Lib → Auth)+2.5
Model Stability (Drift)

Refusal Rate Change

0.0%

Difference over the testing period.

Start: 37.42%End: 37.42%
Paternalism Audit

Persona Refusal Rate

30.5%

Refusals for sensitive user personas.