VideoOdyssey-V - Temporal Grounding: 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 locating when an event happens; higher is better.

Metric: Multiple-choice accuracy (%) on the 65 temporal grounding questions (locate the specific timestamp in the video where a described visual event occurs) 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 May 2028. 18 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3.1 Pro (Preview)49.2#99
2Seed 2.0 Pro47.7#222
3Kimi K2.546.2#266
4GPT-5.244.6#185
5VideoLLaMA3-7B43.1#2050
6Gemini 2.5 Pro40#276
7Video-KTR-7B40
8Qwen 3 VL 235B A22B38.5#485
9InternVL3.5-38B36.9#1021
10Qwen 3 VL 32B36.9#622
11Video-R136.9#1379
12LLaVA-OneVision-1.5-8B35.4#1864
13Kimi-VL-A3B32.3#2340
14Phi-4 Multimodal Instruct29.2#4154
15LLaVA-NeXT-Video-DPO (34B)26.2#2489

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-temporal-grounding · How It Works · Data refreshed daily, snapshot 2026-11-21.