Agents on Rails - Atomic Task Score: leaderboard
Partial-credit hidden-test scores over 19 atomic Rails tasks, using the fixed lemans/miniswen agent. Runs are averaged within each task, then tasks receive equal weight; each model retains its explicit reasoning effort. This October 2026 protocol is distinct from the legacy binary-success board.
Metric: Mean Task Score (%). Source: rubyonrails.org. Saturation forecast: Not forecast. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 4 Argon (High) | 81.16 | #7 (Gemini 4 Argon) |
| 2 | Grok 4.7 (High) | 77.11 | #27 (Grok 4.7) |
| 3 | DeepSeek V4.1 Flash (Max) | 73.26 | #36 (DeepSeek V4.1 Flash) |
| 4 | GPT-6.1 Sol (xHigh) | 72.37 | #8 (GPT-6.1 Sol) |
| 5 | Kimi K3 (Max) | 71.84 | #30 (Kimi K3) |
| 6 | GPT-6 Astra (xHigh) | 70.79 | #5 (GPT-6 Astra) |
| 7 | GPT-6 Luna (Max) | 70.63 | #52 (GPT-6 Luna) |
| 8 | Claude Opus 5.5 (xHigh) | 69.21 | #3 (Claude Opus 5.5) |
| 9 | GLM-5.3 Flash (Max) | 68.37 | #66 (GLM-5.3 Flash) |
| 10 | GPT-6 Astra (Medium) | 66.05 | #5 (GPT-6 Astra) |
| 11 | Muse Spark 1.3 (Max) | 65.53 | #25 (Muse Spark 1.3) |
| 12 | Claude Opus 5.5 (Medium) | 62.74 | #3 (Claude Opus 5.5) |
No result here: #9 Claude Fable 5.1, #10 Claude Sonnet 5.5.
Interactive version: theaggregate.ai/benchmark?slug=agents-on-rails-atomic-task-score · How It Works · Data refreshed daily, snapshot 2026-10-16.