Skip to content
Cisent Message us on WhatsAppMessage us

Manthys · generalist

Mind

It reasons, drafts and coordinates work inside Husker. When a fact isn't on record, it says so instead of making it up.

Hallucination

60%

less fabrication than its base model.

Mind Ultra 2 on SimpleQA, from OpenAI: it answers 36.7% of the questions incorrectly; its base, 90.9%.

Against its base model

It makes up less and says “I don't know”.

SimpleQA, from OpenAI, and AA-Omniscience, from Artificial Analysis, measure how much a model makes up when it doesn't know. Without being asked, Mind Ultra 2 says “I don't know” on 59% of SimpleQA; its base, on 1.1%.

  • Its base model
  • Manthys

Made-up answers on SimpleQAMind Ultra 2

90.9%36.7%

Made-up answers on AA-OmniscienceMind Ultra 2

27.3%19.0%

Says “I don't know” without being asked (SimpleQA)Mind Ultra 2

1.1%59.0%
From 0 to 100%. Question by question, against the same base model. Making up less is better; saying “I don't know” when it doesn't know is too.
90.9%Its base model38.0%36.7%Mind Ultra 222.0%
No instructionWith “if you don't know, say so”
Made-up answers on SimpleQA (4,317 paired questions). Training alone achieves what the instruction achieves on the base model: 36.7% vs. 38.0%. Together they drop to 22.0%.

It says “I don't know” when it doesn't know.

Efficiency

3.5×

fewer tokens than its base, with no loss in accuracy.

Mind Lite 2 on our own set of 30 generalist questions: 30 of 30 at 160 tokens per answer; its base, 29 of 30 at 569.

Efficiency

The same answer, with less.

Tokens per answer for each model, against its base. All three get the 30 questions right.

Mind Lite 2 · 3.5 times fewer

Manthys Mind Lite 2160
Its base model569

Mind Pro 2 · 3.1 times fewer

Manthys Mind Pro 2214
Its base model658

Mind Ultra 2 · the one that spends least

Manthys Mind Ultra 282
Its base model84
From 0 to 700 tokens. Our own set of 30 generalist questions with verifiable answers, the same for each model and its base. Mind Ultra 2's base was already efficient: the saving is small, and even so it spends the least.

Three models

Lite, Pro and Ultra.

All three work inside Husker: analysis, drafting and task coordination.

Manthys Mind Lite 2

160tokens

per answer, with 30 of 30 correct: 3.5 times fewer than its base.

Manthys Mind Pro 2

214tokens

per answer, with 30 of 30 correct: 3.1 times fewer than its base.

Manthys Mind Ultra 2

54points

less fabrication than its base on SimpleQA, and the one that spends least: 82 tokens per answer.

Every figure

Mind's tables.

Hallucination

MeasureMind Ultra 2Its base model
Made-up answers · SimpleQA (4,324)36.7%90.9%
Made-up answers · AA-Omniscience (600)19.0%27.3%
Hallucination rate · SimpleQA38.3%98.8%
Hallucination rate · AA-Omniscience23.0%33.4%
Says “I don't know” without being asked · SimpleQA59.0%1.1%

Measured by Cisent on September 28, 2026, with a protocol fixed before measuring, question by question against the same base model. The hallucination rate is made-up answers over everything not answered correctly.

Efficiency

ModelCorrectTokens per answerIts base: correctIts base: tokens
Manthys Mind Lite 230 of 30160.329 of 30568.5
Manthys Mind Pro 230 of 30213.830 of 30658.0
Manthys Mind Ultra 230 of 3081.830 of 3084.1

Our own set of 30 generalist questions with verifiable answers, the same for each model and its base. It supports the token savings, not an edge in accuracy.

Figures measured by Cisent between September 20 and 28, 2026.

Manthys

The other families.

Manthys Sentinel · cybersecurity

It reads the attack like an analyst.

Explore Sentinel

Manthys Nova · the largest scale

It flags its doubt. So you can verify.

Explore Nova

Manthys Reflex · decisions

It decides what deserves the judge's attention.

Explore Reflex

Book a meeting with the Cisent team.

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 meeting

or write to info@cisent.ai