GPT-2 small: continue a text

124M parameters, trained only to predict the next token (Radford et al., 2019). It runs in this browser; the prompt never leaves the machine.

Model not loaded. First load: about 130 MB, cached afterwards.