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

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

Top models

#ModelScore
1Kimi K3 (Max)82
2Kimi K3 (Low)82
3Claude Opus 4.5 (Thinking)78
4Gemini 3 Pro (Preview) (High)76
5DeepSeek V4 Pro (0813) (Reasoning, Max Effort)74
6Claude Sonnet 4.5 (Thinking)72
7Claude Opus 4.5 (Non-reasoning)72
8Kimi K2.7 Code68
9Inkling (xHigh)68
10Kimi K2.666
11Kimi K2.5 (Thinking)66
12Hy364
13DeepSeek V4 Pro (Reasoning, High Effort)64
14Gemini 2.5 Pro58
15GLM-5.1 (Non-reasoning)58

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