Tool directory · Vibe Coding
qwen3-4b-deslop
New this weekLast checked
A Qwen3-4B fine-tune that drops common AI writing tics. The author reports the slop score fell from 121 to 12.5 per 1,000 words. It runs locally with Ollama.
Our Take
qwen3-4b-deslop is a fine-tune of Qwen3-4B-Instruct for people who want a small local model with fewer assistant writing habits. The author reports that on 40 prompts held out of training, the slop score fell from 121 to 12.5 per 1,000 words, and em dashes fell from 20.5 per 1,000 words to zero. The Q4_K_M weights are a 2.5 GB download and run locally with Ollama or llama.cpp. The post says the voice is flat and the model still invents names and places. Source