ShareGPT

Fresh

The original community-scraped corpus of ChatGPT conversations that bootstrapped Vicuna and the entire open-instruction-tuning era.

Type: SFT Dataset
Publisher: Anonymous Community
Capabilities: Instruction Following Multi Turn Dialog
Runtime: jsonl
License: Unknown
Size: ~90k conversations (cleaned variants)
Published: May 2026
Canonical: huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered

Cite

Notes

Only stored in your browser.

Lift evidence

Eval	Tools known to lift	Source paper
MT-Bench	ShareGPT	-
Massive Multitask Language Understanding (MMLU)	ShareGPT	-

Notable models trained on it

Vicuna 7B/13B/33BWizard-Vicunacountless 2023-2024 community models