VisualWebArena: leaderboard

VisualWebArena: Measures browser, desktop, mobile, or GUI agents operating in interactive environments.

Metric: Success Rate (%). Source: jykoh.com. Saturation forecast: Estimated already saturated. 32 models tracked.

Top models

#ModelScoreOverall rank
1Gemini 2.5 Flash54#245
2GPT-552.9#98
3GLM-4 32B48.7#459
4GPT-4o19.78#343
5Gemini 1.5 Pro11.98#445
6GPT-47.25#438
7Gemini 1.5 Flash6.59#618
8Gemini 1.0 Pro6.04#933
9GPT-3.52.2#818
10Mixtral 8x7B1.76#1143
11Llama 2 70B1.1#1201

No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.

Interactive version: theaggregate.ai/benchmark?slug=visualwebarena · How It Works · Data refreshed daily, snapshot 2026-10-20.