AI Safety Index
66.04 / 100
7th of 17 published models
14 of 16 benchmarks · 95.0% of configured weight
Benchmark profile
Out of 100; higher is safer.
| Benchmark | Score visualization | Score out of 100 |
|---|---|---|
| SIM-VAILMental & Emotional Safety | 100.00 / 100 | |
| Spiral-BenchMental & Emotional Safety | 69.67 / 100 | |
| KORAYouth Safety | 76.79 / 100 | |
| PatientSafetyBenchMedical Advice Safety | 88.63 / 100 | |
| HealthBench-HardMedical Advice Safety | 30.34 / 100 | |
| ELEPHANTManipulation | 26.53 / 100 | |
| DarkBenchManipulation | 56.06 / 100 | |
| HumanAgencyBench (Autonomy)Manipulation | 45.42 / 100 | |
| ASK — AI Scam KnowledgeSecurity | 63.80 / 100 | |
| FairMT-BenchBias & Fairness | 76.56 / 100 | |
| ConfAIdePrivacy / Confidentiality | 98.03 / 100 | |
| SimpleQA VerifiedMisinformation | 24.35 / 100 | |
| SYCON-BenchMisinformation | No published score | |
| HumanAgencyBench (Correct Misinformation)Misinformation | 78.84 / 100 | |
| SystemCheck / RealGuardrailsRule Following | 76.98 / 100 | |
| AgentIFRule Following | No published score |
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-haiku-5-5 · configuration ID anthropic.claude-haiku-5-5.medium-f52e143e104e
