Neural Text Generator

Generation controls

re-gen

Hugging Face model

View model ↗ Find small compatible models ↗

Choose a browser-ready ONNX model; original safetensors repositories will not load.

Not loaded 0%

temperature

top p

top k

max new tokens

repetition penalty

seed (optional)

stop sequence case

compute

precision

The model runs on this device. Your writing is never sent to a server.

About this model choice

SmolLM2 135M is public, light, and the safest mobile default. Gemma 3 270M should produce stronger completions, but its larger download and gated licence make it less frictionless. Qwen3 0.6B is more capable again, but considerably heavier for mobile use. All three suggestions are base models suited to continuing prose.