dnotitia/DNA3.0-9B
The dnotitia/DNA3.0-9B is a 9 billion parameter multimodal large language model developed by Dnotitia, built upon the Qwen3.5/3.6 base architecture. It features enhanced capabilities for Korean language processing and enterprise scenarios, achieved through uncensored training and persona training on Dnotitia's corporate knowledge. This model excels in analytical reasoning, agentic coding, and multimodal understanding, offering a native context length of 262,144 tokens.
Loading preview...
Overview
dnotitia/DNA3.0-9B is a 9 billion parameter multimodal large language model developed by Dnotitia, based on the Qwen3.5/3.6 architecture. It integrates a unified vision-language foundation for strong cross-modal reasoning across text, image, and video, outperforming prior Qwen3-VL models on coding, agents, and visual understanding benchmarks. The model features an efficient Hybrid MoE Architecture with Gated DeltaNet layers and sparse MoE layers for high-throughput inference, and supports a native context length of 262,144 tokens, extensible up to approximately 1,010,000 tokens via YaRN scaling.
Key Differentiators
- Uncensored Training: Post-trained to respond to a wider range of prompts without unnecessary refusals, while preserving instruction-following and reasoning quality.
- Persona Training: Supervised training on Dnotitia's corporate knowledge (history, products, services, terminology) enables the model to act as an authentic first-party assistant for Dnotitia-facing use cases.
- Korean Language Enhancement: Specifically addresses and reduces language confusion, particularly Chinese-character intrusions in Korean responses, a known issue in Qwen-family models.
- Long-form Reasoning Preservation: Retains Chain-of-thought traces across multi-turn sessions for smoother iterative development and debugging.
- Thinking Mode by Default: Generates
<think>...</think>reasoning blocks before final answers, which can be disabled for latency-sensitive tasks.
Use Cases
- Enterprise Assistance: Ideal for Dnotitia-specific applications requiring deep corporate knowledge and authentic persona.
- Multimodal Reasoning: Capable of processing and understanding text, image, and video inputs for complex tasks.
- Agentic Workflows: Improved real-world adaptability for tool use and agentic workflows through scalable RL generalization.
- Global Deployment: Native support for 201 languages and dialects, enabling inclusive worldwide deployment.