What is Dnotitia DNA 3.0?

On June 14, 2026, Dnotitia Inc. released DNA 3.0, an enterprise-ready AI language model family built on Alibaba Cloud's Qwen 3.5/3.6 and enhanced through proprietary post-training. The models are available on Hugging Face [citation:2].

Model Lineup

DNA 3.0 includes models ranging from lightweight versions to large-scale MoE architectures [citation:2]:

  • 0.8B, 2B, 4B: Lightweight for edge environments
  • 9B, 27B: Mid-range for general enterprise use
  • 35B-A3B MoE: Large MoE (3B activated)
  • 122B-A10B MoE: Flagship MoE (10B activated)

Key Features

  • Persona Training: Models better reflect company information and product context, enabling AI responses aligned with internal knowledge and service policies [citation:2]
  • Reduced Response Constraints: Post-trained to minimize language-mixing issues when adapting Qwen-based models — improving Korean-language enterprise usability [citation:2]
  • Long-Context Handling: Inherited from previous generations
  • Reasoning Trace Preservation & Tool Use: For AI agent workflows
  • Coding Support: Built-in development assistance

Integration: Seahorse Cloud

DNA models are already integrated into Seahorse Cloud, Dnotitia's AI data platform that converts enterprise documents and unstructured data into AI-ready knowledge assets — enabling semantic search, context-aware answer generation, and AI agent workflows based on enterprise data [citation:2].

Use Cases

  • Enterprise Q&A systems
  • AI agent deployments
  • Korean-language institutional applications
  • Document-to-knowledge conversion
  • Internal workflow automation

Pricing

Open source — available on Hugging Face.

Pros

  • Wide range of model sizes (0.8B to 122B MoE)
  • Enterprise-focused post-training
  • Persona training for brand alignment
  • Reduced language mixing in non-English contexts
  • Integrated with Seahorse Cloud platform

Cons

  • Built on Qwen (China-based base model)
  • Limited adoption outside Korean market
  • MoE models require specialized deployment
  • Documentation primarily Korean

Who Should Use It?

Perfect for: Enterprise AI teams, Korean-language institutions, and organizations needing adaptable open-source LLMs with persona training.

Verdict

DNA 3.0 demonstrates Dnotitia's pragmatic approach to enterprise AI — starting with Qwen, adding post-training for real business needs, and integrating into a data platform. The MoE options from 35B to 122B are notable [citation:2].

Rating: 4.3/5 - Enterprise-ready open LLMs.