Skip to main content

Cybersecurity statistics / Large Language Models

Nova 2 Lite shows 34.05% single-turn ASR but 7.89% multi-turn ASR.

PublisherCisco
ReportProprietary Problems: How Frontier Closed Models Collapse Under Iterative Pressure
Published27 May 2026
TopicsLarge Language Models, Adversarial Attacks, Model Robustness

Published by Cisco in Proprietary Problems: How Frontier Closed Models Collapse Under Iterative Pressure, 27 May 2026. The figure is taken from the report as published; the full methodology is in the source.

View the original report

Related statistics

GPT-5.4 moves from 2.74% single-turn ASR to 24.68% multi-turn ASR, a 9x increase.
Cisco, 31/05/2026
Anthropic Claude-family single-turn ASR ranges from 2.19% to 3.64%, and multi-turn ASR ranges from 11.16% to 16.20%.
Cisco, 31/05/2026
Grok 4.1 Fast in its non-reasoning configuration records a multi-turn ASR of 88.30%.
Cisco, 31/05/2026
In an earlier open-weight evaluation, multi-turn ASR reached 92.78% against Mistral Large-2.
Cisco, 31/05/2026
Multi-turn attack success rate (ASR) ranges from 7.89% to 88.30% across the 15 closed/proprietary flagship models in the cohort.
Cisco, 31/05/2026
Single-turn attack success rate (ASR) ranges from 2.19% to 64.91% across the 15 closed/proprietary flagship models in the cohort.
Cisco, 31/05/2026

Get the newsletter

Weekly cybersecurity statistics by email.