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

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

Top models

#ModelScore
1Kimi K3 (Low)66
2Kimi K3 (Max)64
3Claude Opus 4.5 (Thinking)62
4Gemini 3 Pro (Preview) (High)60
5Claude Opus 4.5 (Non-reasoning)54
6Claude Sonnet 4.5 (Non-reasoning)46
7GPT-5 (Low)44
8DeepSeek V4 Pro (0813) (Reasoning, Max Effort)44
9O342
10GPT-5 (Medium)42
11Kimi K2.5 (Thinking)42
12DeepSeek V4 Pro (Reasoning, Max Effort)42
13GLM-5 (Thinking)42
14GPT-5 Codex (High)42
15Grok 440

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