DeepSeek-V4.1-Flash kept refining a 3D Boeing 747 for hours, user reports

A user says DeepSeek-V4.1-Flash spent hours iterating on a Boeing 747 model, repeatedly inspecting it, spotting defects, zooming in, diagnosing problems and making fixes. They praised its ability to sustain that process and identify what to work on, calling it the best Boeing benchmark result they had seen from an open-source model “by far.” The test used Claude Code with a /goal prompt requesting a realistic aircraft in Three.js.