Release of CCORE GLB-3D-Coder 4B Model

CCORE has announced the release of the final version of GLB-3D-Coder 4B, a specialized large language model designed for the generation of reusable 3D assets. The model, which is a full fine-tune of the Qwen3.5-4B architecture, is available for download in several formats, including full Safetensors weights and GGUF versions (Q4_K_M and Q8_0). The Q4_K_M GGUF version occupies approximately 2.7 GB of storage.

According to a developer announcement on Reddit, the model functions by converting text descriptions into complete Three.js HTML scenes. These scenes include an “Export GLB” feature, allowing users to download assets for use in external software such as Blender, Unity, or Godot. The model’s capabilities extend beyond static objects to include the generation of animations within the exported GLB files.

Technical Capabilities and Scope

The model is designed to generate a wide variety of assets, including buildings, environments, weapons, vehicles, furniture, and laboratory equipment. In developer-provided comparisons, the model was measured against DeepSeek V4.1 Flash, with the developer noting significant detail in character designs despite the model’s relatively small 4-billion parameter size.

While the model is capable of generating animations, some users in the developer community have noted that current outputs may be limited to static assets or may require manual correction for complex tasks, such as assets requiring specific skeletal rigs.

Data and Accessibility

The developer has indicated that the training dataset used for GLB-3D-Coder is scheduled for public release on November 6. The model and its recommended settings are hosted on Hugging Face.

The release has prompted discussion regarding the economic implications for the 3D modeling industry. Some observers on social media suggested that such specialized, small-scale models could significantly reduce development costs in game production, though these projections remain speculative. Additionally, the compact size of the 4B model has raised questions regarding its potential for deployment on mobile hardware, though performance metrics for such use cases have not been provided.