# Source: https://huggingface.co/Qwen/Qwen-Image-2.1 (accessed 2026-09-26)
# Unified text-to-image generation and image editing model: 7B visual
# generation component (32 single-stream DiT layers), native RGBA transparency,
# editing with up to 10 reference images. Weights published under the Qwen
# Research License Agreement. First published on 2026-09-14 (Hugging Face
# createdAt); last updated 2026-09-21. Context follows the qwen-image-2.0 API
# convention on Alibaba's first-party host (8,192 prompt tokens, no token
# output budget); the number of output images is a request parameter, not a
# token limit.
name = "Qwen-Image-2.1"
description = "Unified text-to-image generation and image editing model with native RGBA transparency and reference-image editing"
family = "qwen"
release_date = "2026-09-14"
last_updated = "2026-09-21"
attachment = true
reasoning = false
temperature = true
tool_call = false
open_weights = true
license = "qwen-research"

[limit]
context = 8_192
output = 0

[modalities]
input = ["text", "image"]
output = ["image"]

[[weights]]
label = "Hugging Face"
url = "https://huggingface.co/Qwen/Qwen-Image-2.1"

[[links]]
label = "Blog"
url = "https://qwen.ai/blog?id=qwen-image-2.1"
type = "announcement"

[[links]]
label = "GitHub"
url = "https://github.com/QwenLM/Qwen-Image-2.1"
type = "docs"