Home Compare ParentBench Quiz Methodology Changelog Report

Compare AI Models

See how different AI models stack up on safety metrics

Back to all models

SafetyScore

AI safety ratings for everyone.

Links

Home Compare Models Find Your AI Methodology

Stay Updated

Get notified about new model evaluations.

Scores are based on publicly available benchmarks and are for educational purposes. They do not constitute endorsements or guarantees of safety.

Compare AI Models | SafetyScore

First Model

vs

Second Model

Claude 3.5 Sonnet

Anthropic

91A-

Better

VS

6 - 0

Llama 3.1 405B

Meta

72C-

Category Breakdown

89Better

B+

HonestyDoes it make stuff up?

+11

78

C+

88Better

B+

FairnessDoes it treat people differently?

+13

75

C

98Better

A+

Refusal to HarmCan you trick it into saying dangerous things?

+30

68

D+

90Better

A-

Manipulation ResistanceDoes it try to manipulate you?

+16

74

C

87Better

B+

Privacy RespectDoes it leak personal info?

+17

70

C-

85Better

B

Straight TalkDoes it just tell you what you want to hear?

+13

72

C-

Claude 3.5 Sonnet scores 19 points higher overall and wins in 6 of 6 categories.