Horizon - RLM - Hard: leaderboard
Metric: Task Pass Rate (%). Source: www.orinlabs.ai. 5 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | claude-opus-4.8 | 21.54 |
| 2 | claude-sonnet-4.6 | 13.85 |
| 3 | gpt-5 | 7.69 |
| 4 | claude-haiku-4.5 | 7.69 |
| 5 | gpt-5-mini | 1.54 |
Interactive version: theaggregate.ai/benchmark?slug=horizon-rlm-hard · How It Works · Data refreshed daily, snapshot 2026-10-09.