Arao AI

Qwen-Image-2.1

7.1BText-to-Imageby Qwen (Alibaba) Start a discussion

Open-weight text-to-image model from Qwen (Alibaba), released 2026-09-20, with a 7.1B visual-generation backbone under the Qwen Research licence (non-commercial research and evaluation only). It unifies text-to-image and image editing: a 32-layer single-stream DiT, native RGBA-transparent generation and editing, up to 10 reference images, native 2K output; community GGUF / INT8 builds are available for local use.

8

Quantized builds

2

Benchmark records

Quantized builds

Community and official quantized releases; "linked records" counts only benchmarks tied to that exact release.

Publisher / repoRevisionFormat · quantTotal sizeLinked recordsUpdatedDetails
ViggleRepositorybb26a0f38e5fe6c124aaccc9187a87eed5d9ed13LoRA · v0.2.1-6step-r128648.1 MBNo linked records2026-09-29View build
ViggleRepositorybb26a0f38e5fe6c124aaccc9187a87eed5d9ed13LoRA · v0.2.1-6step-r2563.8 GBNo linked records2026-09-29View build
MesmerTechRepository868fc8b147de20f2c6e6835b694b2c4d9de9b5a6Safetensors · NVFP44.54 GB1 records2026-09-26View build
ViggleRepositoryb77064be8b3f0b1a13c6a212067cb3d281c60c84LoRA · BF16324.1 MBNo linked records2026-09-25View build
abenzerpsRepository40319fb15542f0ad22921e0124a191a8a935a60aGGUF · Q4_K_M4.29 GBNo linked records2026-09-25View build
unslothRepository2c31ccd392b367a6637841a143813320a02dff55GGUF · Q4_K_M3.91 GBNo linked records2026-09-25View build
Comfy-OrgRepository9a44dbdb47cefd046be9c0a13476192f34c8db8esafetensors · INT816.1 GB1 records2026-09-25View build
QwenRepository790c92633540aa0cb11d9abf19eb46d861714758safetensors · BF1630.84 GBNo linked records2026-09-25View build

Benchmark data

2 configurations

HardwareFrameworkQuantGen timeVRAM usedSamplesEvidence
NVIDIA RTX 5090NunchakuNVFP4 rank-1284.8 s18.74 GB1L0 Self-reported
NVIDIA RTX 5060 Ti 16GBComfyUIINT8 ConvRot18.5 s15.44 GB1L0 Self-reported

Community discussion

No discussions yet — start the first one.

Start a discussion