RoadmapBench: leaderboard
Coding agents are increasingly deployed in real software development, where a single version iteration requires months of coordinated work across many files. However, most existing benchmarks focus predominantly on single-issue bug fixes from Python repositories, with coarse pass/fail evaluation outcomes, and thus fail to capture long-horizon, multi-target development at real engineering scale. To
Metric: Resolved (%) (self-reported). Source: benchmarklist.com. Saturation forecast: Around July 2027. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.7 | 39.1 | #49 |
| 2 | Claude Opus 4.6 | 32.2 | #63 |
| 3 | GPT-5.4 | 29.6 | #79 |
| 4 | Gemini 3.1 Pro (Preview) | 20.9 | #58 |
| 5 | GLM-5.1 | 20.9 | #97 |
| 6 | DeepSeek V4 Pro | 18.3 | #102 |
| 7 | Qwen 3.6 Plus | 16.5 | #120 |
| 8 | Kimi K2.6 | 15.7 | #104 |
| 9 | MiMo-V2.5-Pro | 13.9 | #158 |
| 10 | Kimi K2.5 | 11.3 | #144 |
| 11 | Qwen 3.5 397B A17B | 10.4 | #146 |
| 12 | MiniMax-M2.7 | 10.4 | #258 |
| 13 | Seed 2.0 Pro | 5.2 | #118 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=roadmapbench · How It Works · Data refreshed daily, snapshot 2026-10-20.