Qwen 27B on a single RTX 5090 generates 8-scene animation from one prompt

Acceptable-Object390 · reddit · 2026-09-16

The author ran Qwen 27B locally on a single RTX 5090 using the open-source harness Row-Bot, giving it one prompt: create a visually rich animation showing the model's full capability, no image or video generation allowed.
The model auto-loaded a frontend-design skill via the new progressive skills loading feature and produced 105 seconds of autonomous browser animation across 8 scenes: terminal cold boot, streamed word-by-word language synthesis with live token ticker, real-time math visualizations (Euler's identity, Leibniz series converging to π), a pixel-by-pixel Mandelbrot zoom (57,600 points/frame), a 612-entity knowledge graph, a dual-parallel agent orchestration demo (plan→search→read→synth→write→verify), and a 1,500-particle physics sandbox — all computed and rendered locally.
Takeaway: local models can do far more than people assume, given a well-built harness.

More from coding & agent