Fine-tuned Qwen3.5 4B goes from placing cubes to full 3D scenes
teortaxesTex · x · 2026-10-29
Developer @canalCCore2 shared before/after comparisons using the same base model (Qwen3.5 4B, pre-finetuning). The original model couldn't generate meaningful 3D scenes — at best placing a few cubes and spheres roughly where objects should be, with no recognizable geometry. After fine-tuning across 4 checkpoints, the model produces detailed low-poly scenes like container harbors, regional airports, and neighborhood bars. Same prompts used for base model and all checkpoints, making the improvement stark.
Related event: Fine-Tuned Qwen3.5-4B Generates 3D Scenes with Just 4k Samples(4 posts)→
More from Multimodal
- Kling 4.0 video model lands on Cloudflare AI Gateway at $0.157/sec for 720p — ritakozlov · 2026-10-29
- AI-generated Star Wars-style game demo shows a flyable desert world with TIE fighters — jasonkneen · 2026-10-29
- LiquidAI's LFM3-Audio-2B audio model trends on Hugging Face — LiquidAI · 2026-10-29
- Generating pop-art Halloween portraits with Flora and a vibe board — round · 2026-10-29
- Generating K-Pop music videos with a Claude Music Video skill, no MCP needed — AIandDesign · 2026-10-29
- Granary-v2 open speech model expands to 48 languages with open data and pipeline — rdesh26 · 2026-10-29