Product / UX / Safety & Guardrails

Qwen3Guard

Alibaba's Qwen3-based safety moderation model family (0.6B / 4B / 8B) that classifies prompts and responses as safe, controversial, or unsafe across 119 languages, with a token-level streaming variant for real-time moderation. Deployable via vLLM, SGLang, and HF inference.

Qwen3Guard (0.6B/4B/8B), Apache-2.0 open weights; 119 languages; generative + streaming variants. 8B ~74.7K HF downloads/month, 119 likes (June 2026). Verified live June 2026.

Openness

4 medium confidence

4.0

weights: open(Qwen3Guard 0.6B/4B/8B on HF)
data: not-released
license: Apache-2.0(OSI)

Open weights under Apache-2.0 (OSI); training data not released, so open_weights (4) rather than open_source. Unusually broad multilingual coverage (119 languages) plus a streaming variant.

https://huggingface.co/Qwen/Qwen3Guard-Gen-8B recorded 2026-06-29
Apache-2.0; open weights; 0.6B/4B/8B; 119 languages; safe/controversial/unsafe; ~74,669 downloads/month, 119 likes

Adoption

4 high confidence

4.0

8B variant ~74.7K HF downloads/month (June 2026); cumulative across the three sizes is higher. Strong, fast-growing adoption.

https://huggingface.co/Qwen/Qwen3Guard-Gen-8B recorded 2026-06-29
~74,669 downloads/month, 119 likes (June 2026)

Capability

4 medium confidence

4.0

Among the broadest open guardrails on language coverage, with a streaming mode that few others offer.

https://huggingface.co/Qwen/Qwen3Guard-Gen-8B recorded 2026-06-29
119 languages; prompt+response; streaming variant

Unchanged since 2026-06-29 (last edited, not re-checked)