Benchmark
First, Do NOHARM (v2)
Safety benchmarks examine risks in model behaviour, including how models respond to unsafe requests. Scores are percentages. Higher is better.
- Models measured
- 13
- Scores recorded
- 13
- Sources
- 1
- IQ weight
- 1.16
Score distribution
- 66.2%
- Mean
- 65%
- Median
- 7%
- Std dev
- 57.9%
- 10th percentile
- 74.6%
- 90th percentile
- 55.8%
- Min
This benchmark has a weight of 1.16 for chat models, measured over 13 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | Muse Spark 1.1clinical-result-288· Clinical Benchmarks data docs | 79.7% |
| 2 | Claude Opus 5clinical-result-124· Clinical Benchmarks data docs | 74.6% |
| 3 | Kimi K3clinical-result-125· Clinical Benchmarks data docs | 74% |
| 4 | GPT-5.6 Solclinical-result-126· Clinical Benchmarks data docs | 70.1% |
| 5 | GPT 5.5clinical-result-127· Clinical Benchmarks data docs | 70% |
| 6 | GPT 5clinical-result-294· Clinical Benchmarks data docs | 68.6% |
| 7 | Claude Fable 5clinical-result-289· Clinical Benchmarks data docs | 65% |
| 8 | Gemini 3.1 Pro Previewclinical-result-128· Clinical Benchmarks data docs | 62.6% |
| 9 | Gemini 2.5 Proclinical-result-290· Clinical Benchmarks data docs | 61.9% |
| 10 | Qwen3.5 397B A17Bclinical-result-291· Clinical Benchmarks data docs | 61.1% |
| 11 | Kimi K2.6clinical-result-292· Clinical Benchmarks data docs | 59.1% |
| 12 | GLM 5.1clinical-result-1821· Clinical Benchmarks data docs | 57.9% |
| 13 | DeepSeek R1clinical-result-293· Clinical Benchmarks data docs | 55.8% |
Data credits: Clinical Benchmarks (clinicalbenchmarks.ai), CC BY 4.0; fractions converted to percentages; source documents retain their own terms (licence)