Horizon - Hard — leaderboard
Metric: Task Pass Rate (%). Source: www.orinlabs.ai. 24 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | RLM - claude-opus-4.8 | 21.54 |
| 2 | OpenClaw (LCM) - gemini-3.5-flash | 21.54 |
| 3 | OpenClaw (LCM) - gemini-3.1-pro-preview | 20.31 |
| 4 | OpenClaw (LCM) - claude-opus-4.8 | 16.92 |
| 5 | OpenClaw (LCM) - gpt-5.5 | 15.38 |
| 6 | Claude Code - claude-opus-4.8 | 15.38 |
| 7 | OpenClaw (LCM) - deepseek-v4-pro | 14.06 |
| 8 | RLM - claude-sonnet-4.6 | 13.85 |
| 9 | Codex - gpt-5-codex | 13.85 |
| 10 | Hermes - claude-sonnet-4.5 | 13.85 |
Interactive version: theaggregate.ai/benchmark?slug=horizon-hard · How the rankings work · Data refreshed daily, snapshot 2026-07-22.