Ramp SWE-Bench — leaderboard
Ramp Labs benchmark for background coding agents on realistic financial software engineering work, scored by resolved tasks with the mini-SWE-agent harness.
Metric: Resolved (%). Source: labs.ramp.com. Status: saturated. 24 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Fable 5 | 87.34 |
| 2 | GPT-5.5 | 83.54 |
| 3 | Claude Opus 4.7 | 83.54 |
| 4 | GPT-5.6 Sol | 82.28 |
| 5 | GLM-5.2 | 81.01 |
| 6 | Grok 4.5 | 81.01 |
| 7 | Claude Opus 4.6 | 79.75 |
| 8 | Kimi K2.7 Code | 79.75 |
| 9 | Claude Opus 4.8 | 78.48 |
| 10 | GPT-5.6 Terra | 75.95 |
| 11 | Gemini 3.1 Pro (Preview) | 74.68 |
| 12 | Claude Sonnet 5 | 74.68 |
| 13 | GPT-5.4 | 73.42 |
| 14 | Kimi K2.6 | 73.42 |
| 15 | GPT-5.6 Luna | 73.42 |
Interactive version: theaggregate.ai/benchmark?slug=ramp-swe-bench · How the rankings work · Data refreshed daily, snapshot 2026-07-22.