RiskWebWorld - Product Risk Profile: leaderboard

RiskWebWorld (2026) gives web GUI agents e-commerce risk-management tasks from a production risk-control pipeline, run on real, often uncooperative websites within 20 steps. Task success rate over the 332 product risk profile tasks, in percent; higher is better.

Metric: Task success rate (%) over the 332 product risk profile tasks, for web GUI agents in RiskWebWorld's Browser Use style loop (DOM tree, screenshot and history each step, 1920x1080 viewport, at most 20 steps, stopped after three consecutive failed actions), default model settings; exact match for exact fields, an LLM judge for free-text fields; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 15 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-5.249.7#185
2Qwen 3 VL 235B A22B46.7#485
3Gemini 3 Pro39.2#145
4Qwen 3.5 397B A17B36.7#264
5BU-30B36.1
6Qwen 3.5 35B A3B35.5#492
7Qwen 3 VL 30B A3B27.1#784
8Claude Sonnet 4.522.3#261
9Kimi K2.518.1#266
10InternVL3.5-38B6.3#1021
11uitars-1.5-7B3.9#7905
12GLM-4.6V1.8#710
13ShowUI-2B0
14GUI-G1-3B0
15UI-TARS-72B (RiskWebWorld checkpoint unspecified)0

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

Interactive version: theaggregate.ai/benchmark?slug=riskwebworld-product-risk-profile · How It Works · Data refreshed daily, snapshot 2026-11-21.