Model components / Training & synthetic datasets

WildChat-1M

WildChat-1M is a corpus of around 838k real user-chatbot conversations collected by the Allen Institute for AI (Ai2) by offering free access to GPT-3.5/GPT-4 in exchange for consenting to data collection. It spans 74 languages and includes metadata such as timestamps, hashed IPs, geographic region, model used, and moderation scores. It is a widely cited source of naturalistic instruction and chat data.

Verified live 2026-06-22 via primary sources. ODC-BY licensed and reported ungated by the HF API, though Ai2 datasets sometimes show an access notice.

Openness

5 medium confidence

5.0

license: odc-by
gated: false per HF API

ODC-BY licensed and reported ungated by the HF API, though Ai2 datasets sometimes show an access notice.

https://huggingface.co/api/datasets/allenai/WildChat-1M recorded 2026-06-22
gated=false, license=odc-by

Adoption

3 high confidence

3.0

31,364 monthly downloads on Hugging Face, graded on the training-corpus bands.

https://huggingface.co/api/datasets/allenai/WildChat-1M recorded 2026-07-04
downloads field = 31,364 (30-day)

Capability

4 medium confidence

4.0

A sustained cross-recipe SFT ingredient (Tulu 3 2024, OLMo 3 Dolci 2025), but its own ablation is weak/dated and it does not lead comparisons.

https://huggingface.co/datasets/allenai/tulu-3-sft-mixture recorded 2026-07-04
Named source in the Tulu 3 SFT mixture
https://arxiv.org/html/2405.01470v1 recorded 2026-07-04
WildChat's own ablation (WildLlama) is modest vs 2023 baselines

Unchanged since 2026-07-04 (last edited, not re-checked)