AICA-Bench - Emotion Understanding (Basic): leaderboard

AICA-Bench emotion understanding with direct prompts: classify the emotion expressed in or evoked by an image from a fixed category list. Weighted F1 in percent, higher is better.

Metric: Weighted F1 (%) of emotion classification (expressed emotion recognition and evoked emotion prediction from a predefined category list) with basic direct prompts over AICA-Bench (8,086 affective images from nine public emotion datasets with GPT-4o-generated instructions; closed-source models through their APIs at standard settings, open models on A100 GPUs); higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 23 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 2.5 Flash68.05#463
2Gemini 2.0 Flash67.16#723
3Gemini 2.5 Pro66.97#276
4GPT-4o64.44#751
5Qwen-VL-Max (AICA-Bench checkpoint unspecified)64.07
6GPT-4o Mini60.15#1667
7Qwen-VL-Plus (AICA-Bench checkpoint unspecified)60.04
8Qwen 2.5 VL 7B Instruct56.43#1902
9Qwen 2.5 VL 3B Instruct54.58#2713
10Ovis2-16B54.38#1594
11LLaVA-OneVision54.02#2028
12Ovis2-8B53.63#2047
13Qwen 2 VL 7B Instruct53.52#2573
14InternVL3-14B-Instruct52.91#1801
15MiniCPM-o-2.652.73#2423

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

Interactive version: theaggregate.ai/benchmark?slug=aica-bench-emotion-understanding-basic · How It Works · Data refreshed daily, snapshot 2026-11-21.