SKU SFT-2107 · Sold by External
WildChat-1M
Product specifications
| SKU | SFT-2107 |
|---|---|
| Data type | SFT dataset |
| Volume | 837989 conversations |
| Size on disk | 3.36 GB (Parquet) |
| Format | Parquet (Hugging Face) |
| Access model | PUBLIC LICENSE |
| Pricing | Free · open-source license |
| Quality score | — |
| License | ODC-BY 1.0 |
WildChat-1M is a corpus of real, in-the-wild user conversations with ChatGPT (GPT-3.5 and GPT-4), gathered by the Allen Institute for AI via a free public chatbot in exchange for consent. Rows include the full multi-turn conversation plus rich metadata: model, turn count, language, coarse geolocation (state/country), hashed IP, and OpenAI/Detoxify moderation and toxicity/redaction flags. Despite the '1M' branding, the current public (non-gated) train split contains 837,989 conversations.