Qwen3-VL: Vision Language Text Encoder

Qwen3-VL is Alibaba's latest vision-language model for use as a ComfyUI text encoder, available in 4B and 8B variants with bf16, fp8, and nvfp4 quantization.

Q

Qwen3-VL

Vision-LanguageText EncoderOCRAgentOpen Source

Latest vision-language model from the Qwen series (Alibaba Cloud). Delivers comprehensive upgrades across text understanding, visual perception, extended context (256K native, expandable to 1M), and agent interaction. Packaged as ComfyUI text encoder checkpoints in 4B and 8B sizes with bf16, fp8 scaled, and nvfp4 quantization options.

Hugging FaceGitHubBlogChat
DeveloperAlibaba Cloud (Qwen Team)
Release Date2025-07
ArchitectureDense & MoE Vision-Language Transformer
LicenseApache-2.0
Variants4B (Dense), 8B (Dense)
Context Window256K native (expandable to 1M)
QuantizationsBF16, FP8 scaled, NVFP4 (8B only)
CapabilitiesVisual agent, OCR (32 languages), visual coding, spatial perception, long video understanding

Guides and workflows related to this model series.

No articles found.