FCMBench-Video (English) - Temporal Grounding: leaderboard

FCMBench-Video (2026) English-language handheld document videos from financial credit workflows, task temporal grounding: mean temporal IoU (%) between predicted and ground-truth time spans of the target documents. Zero-shot; higher is better.

Metric: Temporal Grounding: mean temporal IoU (%) between predicted and ground-truth time spans of the target documents, on FCMBench-Video's en-US subset (795 composed videos of 8 English document types, 5,362 instances); zero-shot with structured outputs; each model's native video serving path (2 FPS sampling for most open models, 0.5 FPS for Ovis2.5, raw video upload for the API models); higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 9 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Pro (Preview)80.39#119
2Qwen 3.5 27B75.74#375
3Doubao-Seed-1.6-vision72.23#361
4Qwen 3 VL 32B Instruct71.57#606
5Qwen 3 VL 8B Instruct67.13#1029
6Qwen3 Omni 30B A3B Instruct42.18#930
7InternVL3-8B16.12#1782
8Ovis2.5-9B3.69#1225
9Kimi-VL-A3B-Instruct2.96#2253

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

Interactive version: theaggregate.ai/benchmark?slug=fcmbench-video-english-temporal-grounding · How It Works · Data refreshed daily, snapshot 2026-11-21.