SO-Arena v1.0 - Code Understanding: leaderboard

SO-Arena v1.0 Code Understanding domain score on the source-selected1232-task panel. Domain scores average native task credits for static decisions, tool outcomes or web progress, then scale to0–100. The full release has2919tasks; this leaderboard emphasizes the highest difficulty tiers and keeps all untiered sentiment/knowledge tasks. Metric: Native panel task credit mean (%). Higher is better.

Metric: Native panel task credit mean (%). Source: huggingface.co. Saturation forecast: Around February 2027. 20 models tracked. Updated 14 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1GPT-6 SolMedium92.86#26
2GPT-6 LunaMedium85.71#99
3KnowLine-31B-Gen167.14
4Sieve-31B-v365.71
5PPLX-Decider-v1.261.43
6Clef 27B60
7Jev 1.14.058.57
8GLiDE v1.257.14
9Kev 4B55.71
10StartLux-v1.1-4B-Instant55.71
11Kev 27B54.29
12StartLux-v1.1-4B-AutoThink54.29
13PPLX-Decider-Flash-v1.252.86
14Laya (421M)51.43
15SII-Decision-4B51.43

No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.

Interactive version: theaggregate.ai/benchmark?slug=so-arena-v1-0-code-understanding · How It Works · Data refreshed daily, snapshot 2026-11-16.