VideoOdyssey-V - Counting: 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 counting entities or events; higher is better.

Metric: Multiple-choice accuracy (%) on the 307 counting questions (count the occurrences of specific entities in the video, including humans, other objects, scenes and events) 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 October 2028. 18 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Seed 2.0 Pro34.9#222
2Gemini 3.1 Pro (Preview)34.2#99
3Gemini 2.5 Pro31.6#276
4Qwen 3.5 27B30.9#375
5InternVL3.5-38B30.3#1021
6VideoLLaMA3-7B28.7#2050
7GPT-5.228.3#185
8Qwen 3 VL 235B A22B27#485
9Kimi K2.526.7#266
10Kimi-VL-A3B25.7#2340
11Video-R124.4#1379
12Video-KTR-7B24.4
13Qwen 3 VL 32B22.8#622
14LLaVA-OneVision-1.5-8B21.5#1864
15LLaVA-NeXT-Video-7B-DPO18.2#8618

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

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