AA Omniscience - Software Engineering (SWE) - Python: leaderboard
Metric: Accuracy (%). Source: artificialanalysis.ai. 294 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.5 (Thinking) | 74.5 |
| 2 | Kimi K3 (Max) | 71.5 |
| 3 | Gemini 3 Pro (Preview) (High) | 66.5 |
| 4 | Kimi K3 (Low) | 64.5 |
| 5 | Claude Opus 4.5 (Non-reasoning) | 64 |
| 6 | DeepSeek V4 Pro (0813) (Reasoning, Max Effort) | 63.5 |
| 7 | Kimi K2.7 Code | 55.5 |
| 8 | Claude Sonnet 4.5 (Non-reasoning) | 53 |
| 9 | Claude Sonnet 4.5 (Thinking) | 52.5 |
| 10 | Kimi K2.5 (Thinking) | 47 |
| 11 | Hy3 | 46.5 |
| 12 | DeepSeek V4 Pro (Reasoning, High Effort) | 46 |
| 13 | GPT-5 (High) | 45.5 |
| 14 | DeepSeek V4 Pro (Reasoning, Max Effort) | 45 |
| 15 | Inkling (xHigh) | 45 |
Interactive version: theaggregate.ai/benchmark?slug=aa-omniscience-software-engineering-swe-python · How It Works · Data refreshed daily, snapshot 2026-09-05.