VideoOdyssey-V - Causal Reasoning: leaderboard

VideoOdyssey (2026) tests ultra-long video understanding with human-written multiple-choice questions on videos averaging 109 minutes, each annotated with how long a viewer must watch continuously to answer. This board is accuracy in percent on the visual subset questions about inferring causes and consequences; higher is better.

Metric: Multiple-choice accuracy (%) on the 128 causal reasoning questions (infer the underlying causes or resulting consequences of a specific event) of VideoOdyssey-V (human-written questions on 100 videos of 60 to 255 minutes that need visual cues and long continuous viewing; 1,618 in all), 128 frames for proprietary models and each open model's maximum frame setting; higher is better. Source: arxiv.org. Saturation forecast: Around July 2027. 18 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3.1 Pro (Preview)59.4#102
2Seed 2.0 Pro57.8#226
3Kimi K2.557#272
4Gemini 2.5 Pro52.3#280
5GPT-5.251.6#190
6Qwen 3.5 27B48.4#380
7Qwen 3 VL 235B A22B45.3#486
8InternVL3.5-38B42.2#1026
9Video-KTR-7B41.4
10Video-R137.5#1384
11Qwen 3 VL 32B34.4#624
12Phi-4 Multimodal Instruct32.8#4155
13VideoLLaMA3-7B30.5#2064
14Kimi-VL-A3B28.9#2362
15LLaVA-OneVision-1.5-8B27.3#1876

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=videoodyssey-v-causal-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-22.