The AI arena is free today

Open Superagent

VisualWebBench

Paper

Progress Over Time

Interactive timeline showing model performance evolution on VisualWebBench

State-of-the-art frontier
Open
Proprietary

VisualWebBench Leaderboard

2 models
ContextCostLicense
1
Amazon
Amazon
2
Amazon
Amazon
Notice missing or incorrect data?
About this benchmark

What is VisualWebBench?

A multimodal benchmark designed to assess the capabilities of multimodal large language models (MLLMs) across web page understanding and grounding tasks. Comprises 7 tasks (captioning, webpage QA, heading OCR, element OCR, element grounding, action prediction, and action grounding) with 1.5K human-curated instances from 139 real websites across 87 sub-domains.

VisualWebBench is a multimodal benchmark evaluating models on multimodal, frontend development, and vision tasks. LLM Stats tracks 2 models on this benchmark, scored on a 0–1 scale. The current average is 0.8, with the leader at 0.8.

Compare leaders on the best AI for multimodal, best AI for frontend development and best AI for vision leaderboards.

Current leaders

Nova Pro from Amazon currently leads the VisualWebBench leaderboard with a score of 0.797 across 2 evaluated AI models.

1Nova ProAmazon79.7%
2Nova LiteAmazon77.7%

Source paper

Title
VisualWebBench: How Far Have Multimodal LLMs Evolved in Web Page Understanding and Grounding?
Authors
Junpeng Liu, Yifan Song, Bill Yuchen Lin, Wai Lam, and 3 others
Published
Abstract

Multimodal Large Language models (MLLMs) have shown promise in web-related tasks, but evaluating their performance in the web domain remains a challenge due to the lack of comprehensive benchmarks. Existing benchmarks are either designed for general multimodal tasks, failing to capture the unique characteristics of web pages, or focus on end-to-end web agent tasks, unable to measure fine-grained abilities such as OCR, understanding, and grounding. In this paper, we introduce \bench{}, a multimodal benchmark designed to assess the capabilities of MLLMs across a variety of web tasks. \bench{} consists of seven tasks, and comprises 1.5K human-curated instances from 139 real websites, covering 87 sub-domains. We evaluate 14 open-source MLLMs, Gemini Pro, Claude-3 series, and GPT-4V(ision) on \bench{}, revealing significant challenges and performance gaps. Further analysis highlights the limitations of current MLLMs, including inadequate grounding in text-rich environments and subpar performance with low-resolution image inputs. We believe \bench{} will serve as a valuable resource for the research community and contribute to the creation of more powerful and versatile MLLMs for web-related applications.

FAQ

Common questions about the VisualWebBench benchmark and leaderboard.

What is the VisualWebBench benchmark?

A multimodal benchmark designed to assess the capabilities of multimodal large language models (MLLMs) across web page understanding and grounding tasks. Comprises 7 tasks (captioning, webpage QA, heading OCR, element OCR, element grounding, action prediction, and action grounding) with 1.5K human-curated instances from 139 real websites across 87 sub-domains.

What is the VisualWebBench leaderboard?

The VisualWebBench leaderboard ranks 2 AI models based on their performance on this benchmark. Currently, Nova Pro by Amazon leads with a score of 0.797. The average score across all models is 0.787.

What is the highest VisualWebBench score?

The highest VisualWebBench score is 0.797, achieved by Nova Pro from Amazon.

How many models are evaluated on VisualWebBench?

2 models have been evaluated on the VisualWebBench benchmark, with 0 verified results and 2 self-reported results.

Where can I find the VisualWebBench paper?

The VisualWebBench paper is available at https://arxiv.org/abs/2404.05955. The paper details the methodology, dataset construction, and evaluation criteria.

What categories does VisualWebBench cover?

VisualWebBench is categorized under multimodal, frontend development, and vision. The benchmark evaluates multimodal models.

How recent are the VisualWebBench leaderboard results?

The VisualWebBench leaderboard was last updated in August 2026 and currently includes 2 evaluated models.