TrustShiftProbe - Financial Analysis: leaderboard

Metric: Attack success rate (%, lower is better; share of the financial analysis attack sessions whose switched payload succeeds; each model as a ReAct agent connected to one MCP server that behaves benignly during a conditioning phase and then switches to an adversarial payload; 90 cases per domain; temperature 0, 15 tool calls; no defense). Source: arxiv.org. Saturation forecast: Around 2028. 6 models tracked.

Top models

#ModelScore
1GPT-564.4
2Claude Opus 4.868.2
3GPT-4.175.6
4Grok 4.375.6
5Qwen 3.5 Flash76.2
6O4 Mini (2025-04-16)84

Interactive version: theaggregate.ai/benchmark?slug=trustshiftprobe-financial-analysis · How It Works · Data refreshed daily, snapshot 2026-09-29.