LongShOTBench - Reasoning: leaderboard

LongShOTBench (arXiv:2512.16978) Reasoning (RE) category breakdown: weighted criterion-level rubric score (0-100%) on temporal, causal, and cross-modal reasoning over long omni-modal videos; higher is better.

Metric: Weighted Rubric Score (%). Source: longshot.cvmbzuai.com. Saturation forecast: Around November 2027. 102 models tracked. Updated 21 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1Qwen 3 Omni 30B A3B (Thinking)72.44
2Gemini 3.1 Pro (Preview)65.98#99
3Voxtral-Mini-3B-250758.37#3682
4Voxtral-Small-24B-250756.12#1246
5Qwen 3 VL 235B A22B (Thinking)54.62#459
6Qwen 3 VL 32B (Thinking)52.07#573
7Kimi K2.651.5#192
8Vero-Qwen3I-8B50.6
9Nemotron 3 Nano Omni 30B A3BReasoning50.46#1412
10Qwen 3 VL 30B A3B (Thinking)50.09#719
11Gemma 3 27B (IT)48.36#1324
12Qwen 3 VL 8B (Thinking)47.96#910
13Qwen3 Omni 30B A3B Instruct47.43#930
14Gemma 3 12B (IT)47.02#1887
15Intern-S1-mini46.68#1178

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=longshotbench-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-21.