Manthys Sentinel Ultra 2
79.20
average on CTI-MCQ, CTI-RCM, SecEval and CyberMetric, and 72.44 in ATT&CK technique extraction.
Manthys · cybersecurity
It reads the attack like an analyst: logs, vulnerabilities, ATT&CK techniques and the incident report. Without fabricating a fact that must come from a source.
ATT&CK technique extraction
72.44F1
The best figure published by another specialist is 49.1.
Sentinel Ultra 2 on CTI-Bench (CTI-ATE, 60 reports). 49.1: Cisco's Foundation-Sec-8B-Reasoning (arXiv 2601.21051).
Where it shows most
F1 from 0 to 100. 49.1 is the best figure published by another specialist (Cisco, arXiv 2601.21051).
Average of CTI-MCQ, CTI-RCM, CTI-ATE and CVSS, from 0 to 100. For other models, the figures DeepHat publishes, including those for GPT-OSS-120B and Llama 4 Scout, and Llama-Primus's model card.
Against the market
The average of four third-party benchmarks: CTI-MCQ, CTI-RCM, SecEval and CyberMetric. Sentinel Lite 2, at 4B, performs like the 8B models.
Scale of 0 to 100. Cisent's figures measured by Cisent. For other models, the figures Cisco publishes for its reasoning model (arXiv 2601.21051) and those a third party measured with a single harness (arXiv 2510.14113). Each model with its own protocol.
Benchmark by benchmark
Each benchmark in a row and each model as a dot, the others' too.
CTI-MCQthreat intelligence
CTI-RCMvulnerability root cause
SecEvalsecurity knowledge
CyberMetriccybersecurity knowledge
Average of the fourthe ranking's index
Sentinel judges. Husker decides.
Three models
Manthys Sentinel Ultra 2
79.20
average on CTI-MCQ, CTI-RCM, SecEval and CyberMetric, and 72.44 in ATT&CK technique extraction.
Manthys Sentinel Pro 2
69.16
average on CTI-Bench, above Llama-Primus (63.85) and what DeepHat publishes for its V2 (52.30).
Manthys Sentinel Lite 2
87.15
on the 2,000-question CyberMetric, the version the specialists publish on: they score between 84.0 and 85.1.
What it's for
It analyzes security alerts and logs.
It explains each one and maps it to CWE and to ATT&CK techniques.
It drafts the incident report.
Inside Cisent, the model judges and Husker decides.
Sentinel Red Team runs on Manthys Sentinel models trained for red teaming.
Every figure
A dash is a figure its author doesn't publish.
† Measured by a third party with a single harness (arXiv 2510.14113); its CyberMetric is the 2,000-question version.
‡ What Cisco publishes for its reasoning model (arXiv 2601.21051): SecEval on 1,225 questions and the 2,000-question CyberMetric.
i With the evaluation system prompt · p With the production one.
Cisent measures SecEval on all 2,182 questions and CyberMetric on the 500-question version. A dash is a figure its author doesn't publish.
Cisent's figures measured by Cisent between September 19 and October 11, 2026. For other models, only published figures: by their authors (Cisco, arXiv 2601.21051; DeepHat; Llama-Primus's model card) or by a third party with a single harness (arXiv 2510.14113). Each model with its own protocol.
Manthys
Manthys Mind · generalist
Manthys Nova · the largest scale
Manthys Reflex · decisions
Tell us what your organization needs to protect. We'll reply on WhatsApp and set up the meeting.
Message us on WhatsApp and let's book a meetingor write to info@cisent.ai