AA Omniscience - Software Engineering (SWE) - C: leaderboard

Metric: Accuracy (%). Source: artificialanalysis.ai. 294 models tracked.

Top models

#ModelScore
1Kimi K3 (Max)90
2Claude Opus 4.5 (Thinking)85
3Kimi K3 (Low)81
4Gemini 3 Pro (Preview) (High)78
5DeepSeek V4 Pro (0813) (Reasoning, Max Effort)77
6Kimi K2.7 Code76
7Inkling (xHigh)75
8Kimi K2.674
9DeepSeek V4 Pro (Reasoning, Max Effort)74
10Claude Opus 4.5 (Non-reasoning)74
11Claude Sonnet 4.5 (Thinking)73
12Kimi K2.5 (Thinking)70
13DeepSeek V4 Flash (Reasoning, Max Effort)70
14DeepSeek V4 Pro (Reasoning, High Effort)70
15Inkling Small69

Interactive version: theaggregate.ai/benchmark?slug=aa-omniscience-software-engineering-swe-c · How It Works · Data refreshed daily, snapshot 2026-09-05.