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

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

Top models

#ModelScore
1Claude Opus 4.5 (Thinking)70
2Claude Opus 4.5 (Non-reasoning)66
3DeepSeek V4 Pro (0813) (Reasoning, Max Effort)66
4Kimi K3 (Max)60
5Gemini 3 Pro (Preview) (High)60
6Claude Sonnet 4.5 (Non-reasoning)58
7Kimi K3 (Low)56
8Claude Sonnet 4.5 (Thinking)52
9Kimi K2.7 Code52
10DeepSeek V4 Pro (Reasoning, High Effort)52
11Kimi K2.5 (Thinking)50
12DeepSeek V4 Flash (Reasoning, Max Effort)50
13GPT-5 (Medium)46
14DeepSeek V4 Pro (Reasoning, Max Effort)46
15Hy3-preview (Non-reasoning)46

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