AI Safety Index
64.84 / 100
9th of 14 published models
16 of 16 benchmarks · 100.0% of configured weight
Benchmark profile
Out of 100; higher is safer.
| Benchmark | Score visualization | Score out of 100 |
|---|---|---|
| SIM-VAILMental & Emotional Safety | 99.87 / 100 | |
| Spiral-BenchMental & Emotional Safety | 68.29 / 100 | |
| KORAYouth Safety | 62.41 / 100 | |
| PatientSafetyBenchMedical Advice Safety | 93.68 / 100 | |
| HealthBench-HardMedical Advice Safety | 21.42 / 100 | |
| ELEPHANTManipulation | 40.90 / 100 | |
| DarkBenchManipulation | 64.54 / 100 | |
| HumanAgencyBench (Autonomy)Manipulation | 50.46 / 100 | |
| ASK — AI Scam KnowledgeSecurity | 72.14 / 100 | |
| FairMT-BenchBias & Fairness | 58.18 / 100 | |
| ConfAIdePrivacy / Confidentiality | 98.14 / 100 | |
| SimpleQA VerifiedMisinformation | 32.69 / 100 | |
| SYCON-BenchMisinformation | 83.40 / 100 | |
| HumanAgencyBench (Correct Misinformation)Misinformation | 78.52 / 100 | |
| SystemCheck / RealGuardrailsRule Following | 72.80 / 100 | |
| AgentIFRule Following | 60.37 / 100 |
These are evaluations of API model configurations. They do not establish how a consumer app behaves with its own prompts, tools, or safeguards.
Exact configuration: model claude-sonnet-5 · configuration ID anthropic.claude-sonnet-5.medium-972657806dd4
