Mantus
The all-rounder. Strong reasoning at a comfortable speed. Mantus is the default across Ichnus — the model most people meet first, and the one most people settle into.
Best for
- Multi-step reasoning and planning
- Writing, analysis, and research
- Long-document work
- The default experience across Ichnus
Why this tier exists
Mantus is what most people mean when they say "Ichnus." It is the model behind the default chat experience, tuned to be reliably good across writing, research, planning, and analysis without the wait that comes with a frontier model.
We spend most of our product research on this tier, because it is where the largest number of people actually spend their time. Small honesty and reasoning improvements here compound across more conversations than anywhere else in the family.
Benchmarks
Scores on public benchmarks and two internal evaluations. Figures are illustrative — see the note below.
| Evaluation | Mantus | Krus | Mantus | Corus |
|---|---|---|---|---|
| MMLU | 85.6 | 78.9 | 85.6 | 88.2 |
| GPQA (diamond) | 67.8 | 58.2 | 67.8 | 76.4 |
| MATH (competition) | 71.0 | 62.4 | 71.0 | 79.1 |
| HumanEval | 84.1 | 76.5 | 84.1 | 89.3 |
| OversightQA (internal) (ours) | 79.4 | 71.8 | 79.4 | 84.6 |
| Curos-Helpfulness (internal) (ours) | 86.9 | 80.2 | 86.9 | 88.7 |
Benchmark scores are illustrative and invented for this concept project. Curos is not a real organization and these are not measurements of a real model.
Model card
The sections below mirror the model-card format we publish for every release. For Mantus, the release report is published alongside the evaluation suite it passed.
Overview
- Version: Mantus 3, released April 2026
- Context window: 256k tokens
- Availability: available to all Ichnus users, free of charge
- Safety classification: consumer general-purpose assistant, not for high-stakes uses without review
Training and data
Mantus was trained on a curated corpus, with a strong emphasis on removing low-quality and duplicated data. We publish an outline of our data practices in the model documentation and the details behind each release in its evaluation report.
Known limitations
- May confidently answer outside its knowledge — we are working on honesty, not pretending it is solved
- Not a substitute for professional advice in medicine, law, or finance
- Can produce plausible but wrong code; always review generated code
Safety evaluations
Before release, Mantus passed our full evaluation suite — capability, safety, honesty, and red-teaming — as reviewed by the Safety & Evaluations Committee. The public report lists what we tested, what we found, and what we did not test. See our approach to evaluations for how the suite works.
Where to use it
- Handles multi-step reasoning and longer context
- Best for writing, analysis, and everyday planning
- The default model across Ichnus
- 256k-token context window