Horizon — leaderboard
Orin Labs long-horizon agent-learning benchmark with 195 tasks drawn from months of real customer interactions and scored by task pass rate.
Metric: Task Pass Rate (%). Source: www.orinlabs.ai. 24 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | OpenClaw (LCM) - claude-opus-4.8 | 57.95 |
| 2 | OpenClaw (LCM) - gemini-3.5-flash | 57.44 |
| 3 | OpenClaw (LCM) - gpt-5.5 | 56.92 |
| 4 | RLM - claude-opus-4.8 | 55.9 |
| 5 | OpenClaw (LCM) - gemini-3.1-pro-preview | 52.58 |
| 6 | RLM - gpt-5 | 50.77 |
| 7 | RLM - claude-sonnet-4.6 | 49.74 |
| 8 | Codex - gpt-5-codex | 46.15 |
| 9 | Claude Code - claude-opus-4.8 | 45.13 |
| 10 | OpenClaw (LCM) - claude-sonnet-4.5 | 42.05 |
Interactive version: theaggregate.ai/benchmark?slug=horizon · How the rankings work · Data refreshed daily, snapshot 2026-07-22.