LongShOTBench - Visual: leaderboard

LongShOTBench (arXiv:2512.16978) Visual modality slice: weighted criterion-level rubric score (0-100%) on visually grounded questions over long videos; higher is better.

Metric: Weighted Rubric Score (%). Source: longshot.cvmbzuai.com. Saturation forecast: Around 2029. 102 models tracked. Updated 21 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1Qwen 3 Omni 30B A3B (Thinking)62.53
2Gemini 3.1 Pro (Preview)53.2#99
3Voxtral-Mini-3B-250748.79#3682
4Voxtral-Small-24B-250747.23#1246
5Qwen 3 VL 235B A22B (Thinking)46.19#459
6Qwen 3 VL 32B (Thinking)44.86#573
7Vero-Qwen3I-8B44.22
8Kimi K2.643.14#192
9Qwen 3 VL 30B A3B (Thinking)41.77#719
10Qwen 3 VL 235B A22B Instruct40.52#532
11Gemma 3 27B (IT)40.29#1324
12Nemotron 3 Nano Omni 30B A3BReasoning40#1412
13Qwen 3 VL 8B (Thinking)39.89#910
14Intern-S1-mini39.72#1178
15Qwen 3 VL 32B Instruct38.47#606

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

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