SKU SFT-2104 · Sold by External
Infinity-Instruct
Product specifications
| SKU | SFT-2104 |
|---|---|
| Data type | SFT dataset |
| Volume | 7449106 examples |
| Size on disk | 21.1 GB |
| Format | Parquet (Hugging Face) |
| Access model | GATED |
| Pricing | Free · open-source license |
| Quality score | — |
| License | CC BY-SA 4.0 |
Infinity-Instruct is BAAI's large-scale instruction-tuning collection assembled and evolved from open instruction sources spanning code, math, and general/subjective chat. It is released as multiple configs; the headline chat config (7M) holds 7,449,106 rows, with a 3M foundational config (3,463,473) and dated generation sets (Gen/0729 = 1,456,927; 0625 = 659,808; 0613 = 362,330). Access is gated on Hugging Face (auto-approved terms acceptance). Rows carry ShareGPT-style conversations plus ability/label and language annotations.