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. Saturation forecast: Around January 2027. 26 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 | 88.46 | #12 |
| 2 | Claude Fable 5 | 88.46 | #16 |
| 3 | Kimi K3 | 87.18 | #30 |
| 4 | GPT-5.5 | 83.33 | #31 |
| 5 | Claude Opus 4.7 | 83.33 | #49 |
| 6 | GPT-5.6 Sol | 83.33 | #19 |
| 7 | GLM-5.2 | 82.05 | #73 |
| 8 | Grok 4.5 | 82.05 | #48 |
| 9 | Claude Opus 4.6 | 80.77 | #63 |
| 10 | Kimi K2.7 Code | 80.77 | #89 |
| 11 | DeepSeek V4 Flash (0731) | 79.49 | #78 |
| 12 | Claude Opus 4.8 | 78.21 | #37 |
| 13 | GPT-5.6 Terra | 76.92 | #41 |
| 14 | Claude Sonnet 5 | 75.64 | #57 |
| 15 | Gemini 3.1 Pro (Preview) | 74.36 | #58 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=ramp-swe-bench · How It Works · Data refreshed daily, snapshot 2026-10-20.