SEED-Bench — leaderboard

Comprehensive multimodal benchmark with 19K questions across 12 dimensions evaluating image and video understanding.

Metric: Avg Score (%). Source: huggingface.co. Status: saturation imminent. 63 models tracked.

Top models

#ModelScore
1[InternVL-Chat-V1.2-Plus](https://github.com/OpenGVLab/InternVL)70.4
2[Weitu-VL-1.0](https://weitu.ai/)69.2
3[SPHINXv2-1k](https://github.com/Alpha-VLLM/LLaMA2-Accessory/tree/main/SPHINX)67.5
4[GPT-4V](https://openai.com/research/gpt-4v-system-card)67.3
5[Qwen-VL-plus](https://github.com/QwenLM/Qwen-VL/tree/master?tab=readme-ov-file#qwen-vl-plus)66.8
6[SPHINXv1-1k](https://github.com/Alpha-VLLM/LLaMA2-Accessory/tree/main/SPHINX)63.9
7[LLaVA-v1.5-LoRA](https://llava-vl.github.io)62.8
8[llava-v1.5-7b-finetune]()62.8
9[LLaVA-v1.5-13B-LoRA](https://llava-vl.github.io)62.4
10[InfMLLM-13B](https://github.com/mightyzau/InfMLLM)62.3

Interactive version: theaggregate.ai/benchmark?slug=seed-bench · How the rankings work · Data refreshed daily, snapshot 2026-07-22.