Video-IFBench - Selection: leaderboard
Task-gated instruction satisfaction rate on selection instructions requiring the correct video-grounded branch in Video-IFBench (2026). Percent, higher is better.
Metric: Task-gated instruction satisfaction rate (%; TISR on selection instructions, where the model must pick the one true video-grounded condition branch; checklist-based instruction following on 1.5K video samples with 39 semantic and format constraint types; a response counts only if it addresses the active task (the correct branch for conditional instructions) and satisfies every checklist item, judged by Qwen3.5-397B-A17B-Instruct or verified programmatically; Gemini and Doubao get video at 1 fps, GPT-5.4 50 frames, vision-only models frames plus timestamped subtitles). Source: arxiv.org. Saturation forecast: Around February 2027. 38 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 59.2 | #135 |
| 2 | Gemini 3 Flash | 54.8 | #159 |
| 3 | Qwen 3.5 397B A17B (Thinking) | 44.9 | #242 (Qwen 3.5 397B A17B) |
| 4 | Qwen 3.5 122B A10B (Thinking) | 39.4 | #329 (Qwen 3.5 122B A10B) |
| 5 | Doubao-Seed-2.0-Pro-260215 | 38.2 | #149 |
| 6 | Qwen 3.5 27B (Thinking) | 34 | #358 (Qwen 3.5 27B) |
| 7 | Qwen 3.5 9B (Thinking) | 32 | #803 (Qwen 3.5 9B) |
| 8 | Qwen 3.5 35B A3B (Thinking) | 31.5 | #480 (Qwen 3.5 35B A3B) |
| 9 | Gemma 4 31B (IT) | 25.4 | #343 |
| 10 | Qwen 3 VL 235B A22B (Thinking) | 24.9 | #451 |
| 11 | Gemma 4 26B A4B (IT) | 19.4 | #545 |
| 12 | Qwen 3.5 397B A17B (Non-reasoning) | 18.9 | #242 (Qwen 3.5 397B A17B) |
| 13 | Qwen 3.5 27B (Non-reasoning) | 18 | #358 (Qwen 3.5 27B) |
| 14 | Qwen 3 VL 30B A3B (Thinking) | 17.1 | #710 |
| 15 | GPT-5.4 | 16.9 | #121 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=video-ifbench-selection · How It Works · Data refreshed daily, snapshot 2026-11-12.