Open source
research datasets.
We believe in contributing high-quality, verified evaluation sets to the global research community once our models achieve stable compliance.
Vilcus-1 Model Training In Progress
Notice: Vilcus-1 is currently undergoing active training and parameter optimization in our custom compute clusters. The model and its verified performance benchmarks are **not ready yet**. Verified datasets and official benchmarks will be released systematically alongside the public model rollout.
Upcoming Releases
Below are the foundational datasets currently used to train the reasoning alignment phases of Vilcus-1. These will be open-sourced for non-commercial academic research.
LogicReason-100K (Upcoming)
A highly curated dataset of 100,000 multi-turn logical deduction rationales, designed to train models on structural step-by-step thinking.
CodeSolve-50K (Upcoming)
50,000 complex software engineering problems paired with compiler feedback loops, unit tests, and self-correction rationales.
SafeAlign-20K (Upcoming)
20,000 alignment examples focused on structural constraints, safety boundaries, and verifiably low factual hallucination rates.
Get notified on model launch.
Be the first to access Vilcus-1 weights, API endpoints, and our open-source research datasets.
Join the Waitlist