CaST-Bench (Full Set) - Temporal Grounding: leaderboard

Temporal grounding of the evidence chain on the full CaST-Bench set (2026) of 2,066 six-option video questions over 1,015 videos, where the model first outputs a grounded spatio-temporal causal chain and then answers: mean temporal IoU of predicted evidence segments matched one-to-one to the annotated ones, in percent. Higher is better.

Metric: IM-tIoU (%): mean temporal IoU of predicted evidence segments greedily matched one-to-one to the annotated causal-chain instances, unmatched instances counting zero, all 2,066 six-option questions over 1,015 videos; every model must first output a grounded spatio-temporal causal chain (evidence segments with per-second boxes) and then answer; native video input except the GPT-5 models, which get 1 fps frames; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 13 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 2.5 Flash27.63#463
2GPT-526.61#178
3Gemini 2.5 Pro21.53#276
4GPT-5 Mini19.89#364
5GLM-4.1V-9B (Thinking)16.58
6Qwen 3 VL 4B Instruct11.21#1332
7Qwen 3 VL 8B Instruct10.51#1029
8Qwen 3 VL 8B (Thinking)10.23#910
9MiMo-VL-7B-RL-25089.63#823
10InternVL3.5-14B8.77#1504
11InternVL3.5-8B8.59#1604
12InternVL3.5-30B-A3B8.33#1267
13Qwen 2.5 VL 7B Instruct3.72#1902

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

Interactive version: theaggregate.ai/benchmark?slug=cast-bench-full-set-temporal-grounding · How It Works · Data refreshed daily, snapshot 2026-11-21.