SO-Arena v1.0 - Safety: leaderboard
SO-Arena v1.0 Safety domain score on the source-selected1232-task panel. Domain scores average native task credits for static decisions, tool outcomes or web progress, then scale to0–100. The full release has2919tasks; this leaderboard emphasizes the highest difficulty tiers and keeps all untiered sentiment/knowledge tasks. Metric: Native panel task credit mean (%). Higher is better.
Metric: Native panel task credit mean (%). Source: huggingface.co. Saturation forecast: Around February 2027. 20 models tracked. Updated 14 Nov 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Sieve-31B-v3 | 78.48 | ||
| 2 | KnowLine-31B-Gen1 | 77.22 | ||
| 3 | GPT-6 Sol | Medium | 75.32 | #26 |
| 4 | PPLX-Decider-v1.2 | 74.68 | ||
| 5 | GLiDE v1.2 | 74.68 | ||
| 6 | GPT-6 Luna | Medium | 70.25 | #99 |
| 7 | Jev 1.14.0 | 68.99 | ||
| 8 | JevAny-27B | 65.19 | ||
| 9 | Kev 27B | 63.92 | ||
| 10 | H2O-Lightning-4B | 63.92 | ||
| 11 | StartLux-v1.1-4B-Instant | 63.29 | ||
| 12 | StartLux-v1.1-4B-AutoThink | 62.03 | ||
| 13 | PPLX-Decider-Flash-v1.2 | 60.76 | ||
| 14 | Clef 27B | 59.49 | ||
| 15 | Jeva-4B | 50.63 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=so-arena-v1-0-safety · How It Works · Data refreshed daily, snapshot 2026-11-16.