GameplayQA - Timestamp Referring: leaderboard
Metric: Accuracy (%) on the 81 Timestamp Referring questions (the model must identify what exists in a given time range) of GameplayQA (multiple-choice questions on synchronized first-person 3D gameplay videos with structured distractors, zero-shot; Gemini models take the whole video, the others 1 frame per second up to 32 frames); higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash | 80.2 | #93 |
| 2 | Gemini 2.5 Pro | 77.8 | #145 |
| 3 | Qwen 3 VL 30B A3B Instruct | 77.8 | #365 |
| 4 | Qwen 3 VL 235B A22B Instruct | 76.5 | #264 |
| 5 | Qwen 3 VL 8B Instruct | 75.3 | #401 |
| 6 | GPT-5 | 70.4 | #91 |
| 7 | Gemini 2.5 Flash | 69.1 | #237 |
| 8 | Seed 1.6 Flash | 67.9 | #427 |
| 9 | GPT-5 Mini | 66.7 | #176 |
| 10 | GPT-5 Nano | 65.4 | #415 |
| 11 | Seed-1.6 | 65.4 | #257 |
| 12 | Gemma 3 27B (IT) | 64.2 | #509 |
| 13 | Gemma 3 12B (IT) | 54.3 | #655 |
| 14 | Gemma 3 4B (IT) | 54.3 | #971 |
| 15 | Claude Haiku 4.5 | 53.1 | #271 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=gameplayqa-timestamp-referring · How It Works · Data refreshed daily, snapshot 2026-10-11.