WNC LogoWNC
ServicesSolutionsEntrepriseAcadémieLabsActualitésContact
Partenaire
Ouvrir Kanban
WNC LogoWNC

Partenaire IA & automatisation. Conçu avec soin en Ukraine, au Brésil et au Canada.

daniel@we-nocode.com+38 (068) 758-94-22+38 (095) 411-95-37
100 % à distance, partout dans le monde

Services

  • Agentic AI Systems
  • AI-Native Software
  • AI Cloud Infrastructure
  • Consulting & Advisory
  • AI Training for Companies
  • AI Agents & Chatbots
  • Workflow Automation
  • RAG Pipelines
  • Performance Analytics
  • Web Development
  • Mobile Development
  • Desktop Development
  • Server Development
  • Embedded Firmware
  • Embedded Linux & SoC
  • Edge AI & TinyML
  • Wireless & Connectivity
  • Hardware Prototyping
  • IoT Solutions

Solutions

  • Atflow
  • Alt Auth
  • AdCrier
  • Pastukh
  • Olymp Hill

Portails

  • Portail client
  • Portail investisseurs
  • Portail partenaires

Entreprise

  • À propos
  • Notre équipe
  • Témoignages
  • Investissement
  • Blog
  • Actualités
  • Carrières
  • Implantations
  • Contact

Labs

  • Miltech
  • AI Labs
  • Internet of Things
  • Robotics

Initiatives

  • Startup Program
  • Academy
  • Olymp Hill

© 2026 WNC (WeNoCode). Tous droits réservés.

🇺🇦Soutenez l’Ukraine
Back to News
WNC Is Building Its Own Code Model: Introducing Otara-v0
July 4, 2026Company Announcement

WNC Is Building Its Own Code Model: Introducing Otara-v0

Building our own model, from the ground up

Today we're sharing that WNC is developing Otara-v0, our first in-house model focused on code. Rather than fine-tuning someone else's system, we're building the entire stack ourselves — the architecture, the training pipeline, and the tokenizer are all our own. Otara-v0 starts from zero, with no third-party pretrained weights. It's an early, foundational step, and we're being open about the fact that this is a v0: a first working version we intend to grow deliberately.

Trained responsibly on open source

Otara-v0 learns from public code, but "public" isn't the same as "free to use," and we treat that distinction seriously. Our training data is drawn only from repositories under permissive open-source licenses such as MIT, Apache-2.0, BSD, and ISC. Everything goes through a strict pipeline: license filtering, near-duplicate removal, cleaning, and the scrubbing of secrets and personal data. We'd rather start smaller and cleaner than fast and unclear.

Why build it ourselves

Owning the model end to end means owning the decisions — how it's trained, what data shapes it, and how it behaves. For a tool meant to help people write software, that control matters. It lets us understand exactly what's inside Otara, iterate on our own terms, and build something we can stand behind rather than a black box we merely rent.

"Otara-v0 is the start of a long road, not the finish line. We're building it ourselves because we want to know exactly what our tools are made of — and to prove you can train a capable code model on a clean, responsibly sourced foundation."

— Daniel Nikulshyn, CEO

Our ambition

We're not being shy about where we want this to go. Our goal is for Otara to become one of Europe's leading independent models for coding first — built and owned here, on the continent, rather than licensed from elsewhere. We're aiming high: a model whose quality can stand alongside the best in the industry, including the frontier systems people know today. Otara-v0 is only the first step toward that, and we know the bar is high. That's exactly why we're starting from clean foundations and building it ourselves.

Back to all news