KORA V2 - Child Safety: leaderboard
KORA V2 (2026): child and teen safety in conversations, scored against the published V2 rubric. Versioned safety behavior, not general knowledge or reasoning accuracy. Exact public model labels retained; no merge with V1 or unreleased V3. Safety Score (%).
Metric: Safety Score (%). Source: korabench.ai. Saturation forecast: Around January 2027. 42 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra | 73 | #6 |
| 2 | Claude Haiku 4.5 | 70 | #280 |
| 3 | GPT-5.6 Sol | 69 | #18 |
| 4 | Claude Sonnet 4.6 | 67 | #91 |
| 5 | GPT-5.5 | 67 | #30 |
| 6 | Claude Opus 4.6 | 66 | #62 |
| 7 | Claude Opus 4.8 | 65 | #37 |
| 8 | Claude Sonnet 4.5 | 64 | #141 |
| 9 | GPT-5.2 | 61 | #110 |
| 10 | Claude Opus 4.5 | 60 | #83 |
| 11 | GPT-5.4 | 59 | #80 |
| 12 | Claude Opus 4.7 | 57 | #48 |
| 13 | Claude Fable 5.1 | 57 | #9 |
| 14 | Claude Fable 5 | 54 | #15 |
| 15 | Claude 3.5 Haiku | 54 | #564 |
No result here: #4 Claude Opus 5.5, #7 Gemini 4 Argon, #8 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=kora-v2-child-safety · How It Works · Data refreshed daily, snapshot 2026-10-21.