Artifact-Bench (AI-Generated Video) - Real vs AI-Generated Classification: leaderboard

Artifact-Bench (2026) tests whether multimodal LLMs perceive the artifacts of AI-generated videos, on 1,100 manually annotated questions over real videos and clips from seven video generators, split into three difficulty levels. This task asks whether a single video is real or AI-generated; accuracy is the mean of the three difficulty levels, in percent, higher is better.

Metric: Accuracy (%) on the 500 binary real versus AI-generated video questions (RVAC), the unweighted mean of the three difficulty levels, 1,100 manually annotated questions on 1,350 real and AI-generated videos from seven generators, at 5 fps frame sampling, each model at its officially recommended decoding (Gemini 3.1 Pro at temperature 1.0 and high thinking level) or greedy, answers parsed by Gemini 3.1 Pro; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 19 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3.1 Pro (Preview)74#102
2VideoVeritas-8B68.2
3Gemini 3 Flash64.7#180
4Qwen 3 VL 32B Instruct53.5#607
5BusterX++ 7B52
6Skyra-7B-RL51.4
7Qwen 3 VL 32B (Thinking)50.9#577
8MiMo-VL-7B-RL50#1160
9MiMo-VL-7B-SFT49.7#1097
10Qwen 3 VL 8B Instruct49.6#1043
11Qwen 3 VL 30B A3B (Thinking)49.3#727
12Skyra-7B-SFT49.2
13Qwen 3 VL 30B A3B Instruct48.9#903
14Qwen 3 VL 8B (Thinking)48.8#924
15Keye-VL-1.5-8B48.7#1729

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=artifact-bench-ai-generated-video-real-vs-ai-generated-classification · How It Works · Data refreshed daily, snapshot 2026-11-22.