Agents on Rails - Multi-step Task Score: leaderboard
Mean partial-credit score on seven long-horizon tasks in new and existing Rails apps using lemans/miniswen. Required-check failures earn zero; optional checks add credit. Task means have equal weight, with run counts and reasoning effort retained.
Metric: Mean Task Score (%). Source: rubyonrails.org. Saturation forecast: Around April 2027. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5.5 (xHigh) | 72.14 | #3 (Claude Opus 5.5) |
| 2 | Claude Opus 5.5 (Medium) | 60.57 | #3 (Claude Opus 5.5) |
| 3 | GPT-6 Astra (xHigh) | 52.14 | #5 (GPT-6 Astra) |
| 4 | GPT-6.1 Sol (xHigh) | 52.14 | #8 (GPT-6.1 Sol) |
| 5 | GPT-6 Astra (Medium) | 26.86 | #5 (GPT-6 Astra) |
| 6 | Gemini 4 Argon (High) | 26.14 | #7 (Gemini 4 Argon) |
| 7 | Kimi K3 (Max) | 23.29 | #30 (Kimi K3) |
| 8 | Grok 4.7 (High) | 14.71 | #27 (Grok 4.7) |
| 9 | DeepSeek V4.1 Flash (Max) | 7.43 | #36 (DeepSeek V4.1 Flash) |
| 10 | GPT-6 Luna (Max) | 5.71 | #52 (GPT-6 Luna) |
| 11 | GLM-5.3 Flash (Max) | 4.29 | #66 (GLM-5.3 Flash) |
| 12 | Muse Spark 1.3 (Max) | 0 | #25 (Muse Spark 1.3) |
No result here: #9 Claude Fable 5.1, #10 Claude Sonnet 5.5.
Interactive version: theaggregate.ai/benchmark?slug=agents-on-rails-multi-step-task-score · How It Works · Data refreshed daily, snapshot 2026-10-16.