Local test: Nvidia Nemotron 120B crushes Qwen 27B at HTML animation storytelling
nullrecord · reddit · 2026-10-20
A Reddit user compared two local models on the same prompt — a single-file HTML canvas animation explaining noise-cancelling headphones with frosted-glass UI — on an AMD Strix Halo 128GB box running LM Studio.
- Nvidia Nemotron 3.5 Super VL 120B A12B (Q4KM): 7 minutes of thinking at 31 tps, produced a clearly superior animation
- Qwen 3.8 27B (Q80): 20 minutes of thinking at 15 tps, notably worse result
The author concludes the Nemotron model 'hits way above its weight class', with imgur links showing both outputs.
More from coding & agent
- Skyrim AI mod lets bards compose songs about your adventures — and got flamed off Reddit — ReignMocap · 2026-10-20
- LangChain's Harrison Chase: how vertical agents differentiate via context, harness and model — hwchase17 · 2026-10-20
- Coding agents' video skills explode — sell shovels for the vibe-knowledge wave, not videos — AlchainHust · 2026-10-20
- DeepSeek helps root and de-cloud a baby camera, verifying across two reboots — IgorCarron · 2026-10-20
- Cloudflare cut its AI code reviewer's inference bill by 68% with prompt caching and model routing — DanielLockyer · 2026-10-20
- Hermo: open-source tool turns your ChatGPT tab into a callable API with MCP support — MKP_Nimilka · 2026-10-20