SO-Arena v1.0 - Null Discrimination: leaderboard

SO-Arena v1.0 native null discrimination. Stability measures consistency under option reordering, null discrimination measures missing-answer handling, and reliability combines the two. The published JSON supplies these exact values; StartLux null scores are derived from rounded reliability values. Native model inference paths and API interfaces remain separate. Metric: Native null discrimination (0-1). Higher is better.

Metric: Native null discrimination (-1 to 1). Source: raw.githubusercontent.com. Saturation forecast: Rough model projection: around 2026. 19 models tracked. Updated 14 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1JevAny-27B0.7934
2Jev 1.14.00.7794
3Kev 27B0.764
4StartLux-v1.1-4B-AutoThink0.7555
5KnowLine-31B-Gen10.7104
6Sieve-31B-v30.7069
7StartLux-v1.1-4B-Instant0.6991
8GLiDE v1.20.6384
9Clef 27B0.6319
10PPLX-Decider-v1.20.6134
11JevAny-4B0.6024
12Clef flash 9.5B0.5913
13GPT-6 LunaSO-Arena run: Decisions API; model=gpt-6-luna0.55#99
14PPLX-Decider-Flash-v1.20.508
15Kev 4B0.5036

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