A user tested the "oak tree Blender challenge" locally using Qwen3.8-Flash-Next NVFP4 quantized model via the Strata inference engine on an RTX 5090 32GB GPU. The setup featured a 256K token context window running on an AMD Ryzen 9 9950X3D with 64GB DDR5-6000 RAM under Windows. This local execution aimed to replicate a benchmark originally comparing cloud models Opus 5.5 and GPT-6.1 Terra.