use case · basics
The core of autoregressive generation: type a prompt, and GPT-2 keeps writing. Tokens stream in as the model produces them — the same left-to-right, one-token-at-a-time process every LLM uses.
This uses greedy decoding — GPT-2 always takes its single most likely next token — so the output is deterministic. Head to Practical to see the ranked alternatives it passed over, or Wild to steer them yourself.
← Back to GPT-2 · Next: Practical →