ReplaySCM - Ordered: leaderboard
Metric: HeldoutWorldExact (mean share of held-out intervention worlds replayed exactly, 0-1; 250 problems). Source: arxiv.org. Saturation forecast: Around June 2027. 9 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.4 (Medium) | 0.73 |
| 2 | Gemini 3.1 Pro (Preview) (Medium) | 0.64 |
| 3 | Grok 4.20 0309 (Reasoning) | 0.53 |
| 4 | Grok 4 | 0.52 |
| 5 | Grok 4.3 | 0.24 |
| 6 | Kimi K2 (Thinking) | 0.16 |
Interactive version: theaggregate.ai/benchmark?slug=replayscm-ordered · How It Works · Data refreshed daily, snapshot 2026-09-25.