AA Omniscience - Software Engineering (SWE) - R: leaderboard
Metric: Accuracy (%). Source: artificialanalysis.ai. 294 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.5 (Thinking) | 64 |
| 2 | Kimi K3 (Max) | 64 |
| 3 | DeepSeek V4 Pro (0813) (Reasoning, Max Effort) | 64 |
| 4 | Claude Opus 4.5 (Non-reasoning) | 60 |
| 5 | Kimi K3 (Low) | 60 |
| 6 | Claude Sonnet 4.5 (Thinking) | 56 |
| 7 | Gemini 3 Pro (Preview) (High) | 50 |
| 8 | Kimi K2.7 Code | 44 |
| 9 | Gemini 2.5 Pro | 42 |
| 10 | O3 | 42 |
| 11 | GPT-5.1 (High) | 42 |
| 12 | Inkling (xHigh) | 42 |
| 13 | Grok 4 | 40 |
| 14 | Kimi K2.5 (Thinking) | 40 |
| 15 | MiMo-V2.5-Pro (Non-reasoning) | 40 |
Interactive version: theaggregate.ai/benchmark?slug=aa-omniscience-software-engineering-swe-r · How It Works · Data refreshed daily, snapshot 2026-09-05.