Three-tiered severity classification
According to the model card, the Gen variant categorizes outputs into safe, controversial, and unsafe severity levels for adaptation to diverse deployment scenarios.
Open Source Model Profile · Qwen
Qwen3Guard-Gen-4B
Qwen3Guard-Gen-4B is a 4.41B-parameter Qwen3 safety-moderation model from Qwen. According to the model card, this generative Gen variant frames safety classification as instruction following with three-tiered severity levels.
- Publisher
- Qwen
- Task
- text-generation
- Model type
- qwen3
- License
- apache-2.0
- Library
- transformers
- Publication status
- Approved for indexing
Model overview
Qwen3Guard-Gen-4B is published by Qwen as a Qwen3 text-generation model for safety moderation. The captured configuration identifies Qwen3ForCausalLM and Safetensors metadata reports 4,411,424,256 parameters. According to the model card, Qwen3Guard is built upon Qwen3 and trained on 1.19 million safety-labeled prompts and responses, with this repository hosting the generative Gen variant alongside a Stream variant for real-time monitoring.
Recorded capabilities
119-language support
According to the model card, Qwen3Guard-Gen supports 119 languages and dialects for global and cross-lingual use.
Generative instruction-following design
According to the model card, the Gen variant frames safety classification as an instruction-following task, distinct from the token-level Stream variant.
Documented moderation workflows
According to the model card, Transformers and OpenAI-compatible serving examples show prompt and response moderation with Safety and Categories output labels.
Use cases in the source record
- Prompt moderation that classifies user inputs into safe, controversial, or unsafe levels with harm categories, per the documented Transformers workflow.
- Response moderation that assesses generated replies, as shown in the model card's response-moderation and OpenAI-compatible serving examples.
Limitations and unknowns
- Publisher-described benchmark leadership is a source claim and was not independently verified from this record.
- No context-window value was extracted from this record.
- Provider state is historical snapshot data, not independently refreshed current availability.
Source and provenance
Source: Qwen/Qwen3Guard-Gen-4B
Captured: Unknown. Processed: 2026-09-07T19:35:51.397522+00:00.
Qwen3Guard-Gen-4B Qwen3Guard is a series of safety moderation models built upon Qwen3 and trained on a dataset of 1.19 million prompts and responses labeled for safety. The series includes models of three sizes (0.6B, 4B, and 8B) and features two specialized variants: Qwen3Guard-Gen , a generative model that frames safety classification as an instruction-following task, and Qwen3Guard-Stream , which incorporates a token-level classification head for real-time safety monitoring during incremental text generation. This repository hosts Qwen3Guard-Gen , which offers the following key advantages: Three-Tiered Severity Classification: En…
F001F002F003F004F005F006F007F008F009F010F011F012F013F014F015F017F019F020