Magpie
Fresh
A self-synthesis method (and family of datasets) that elicits high-quality instructions directly from an aligned LLM using only its chat template - no seed prompts required.
- Type
- SFT Dataset
- Publisher
- Magpie Align
- Capabilities
- Instruction FollowingMulti Turn Dialog
- Runtime
hf_parquet- License
- Llama-3 Community / Apache-2.0 (varies by split)
- Size
- 1M+ rows (multiple splits, Magpie-Pro is 1M)
- Published
- May 2026
Cite
Notes
Only stored in your browser.
Lift evidence
3| Eval | Tools known to lift | Source paper |
|---|---|---|
| AlpacaEval | Magpie | - |
| Arena-Hard | Magpie | - |
| MT-Bench | Magpie | - |
Models
Notable models trained on it
Llama-3-8B-Magpie-AlignMagpie-Align variants on Mistral, Qwencomponent of Tülu 3 SFT mixture