FeatureArc L4 - Prefix Depth: leaderboard
FeatureArc tests feature development across successive stages in the same codebase. Results measure completed stages, complete features and uninterrupted progress. Mean fraction of consecutively passed stages (%).
Metric: Mean fraction of consecutively passed stages (%). Source: arxiv.org. Saturation forecast: Around February 2028. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5.5 | 64.9 | #3 |
| 2 | GPT-6 Astra | 50.4 | #5 |
| 3 | GLM-5.3 | 45.8 | #42 |
| 4 | Gemini 3.8 Flash | 42 | #25 |
| 5 | Grok 4.7 | 40.9 | #27 |
| 6 | Kimi K3 | 35.7 | #35 |
| 7 | DeepSeek V4 Pro | 31.6 | #109 |
No result here: #7 GPT-6.1 Sol, #9 Gemini 4 Argon, #10 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=featurearc-l4-prefix-depth · How It Works · Data refreshed daily, snapshot 2026-10-29.