AI·THE EXPLAINERニュースを理解する
Alibaba releases open-weight Qwen-Image-2.1 image modelアリババ、画像生成・編集モデルQwen-Image-2.1を公開
Qwen-Image-2.1 combines image generation and editing. Its visual generation component has 7 billion parameters; comparative performance claims come from Qwen’s own benchmarks.アリババのQwenチームが、画像生成と編集を統合したモデルを公開しました。70億パラメータは画像生成部分の規模です。他モデルを上回るという評価は同社のベンチマークに基づく主張です。
Up to 10 references参照画像は最大10枚
Generate + edit生成と編集を統合
Native RGBA透明背景に対応
SIGNAL explanatory diagram · Not a product screenshotSIGNAL解説図 · 実際の製品画面ではありません
SIGNAL diagram of the capabilities described by Qwen. The shapes are illustrative; they are not model-generated samples. 7B refers to the visual generation component.Qwenの公式説明をもとにSIGNALが作図。図形は説明用で、モデルの生成結果ではありません。7Bは画像生成部分のパラメータ数です。
Source for the diagram ↗図の根拠となる公式資料 ↗THE SOURCE TRAIL情報の根拠
1 first-party source linked. 0 news organizations linked.当事者・公的資料1件、報道機関0件にリンクしています。
Inspect the sources ↓出典を確かめる ↓Editorial correction · 2026-09-22編集上の訂正 · 2026-09-22
Clarified that 7B describes the visual generation component; attributed performance comparisons to Qwen.70億パラメータは画像生成部分の規模と明確化。性能比較はQwen自身の評価であると明記しました。
What happened何が起きたか
Alibaba's Qwen AI team released Qwen-Image-2.1, an open-weight model combining text-to-image generation and image editing in one workflow. Its visual generation component has 7 billion parameters and, according to Qwen's own benchmark, outperforms most closed-source models, though independent benchmarks are still pending. The model natively supports transparent (RGBA) image generation and editing, local edits guided by circles, masks or painted marks, and composition using up to ten reference images for uses like group portraits, virtual try-ons or room design. It supports native 2K output across multiple aspect ratios, runs on consumer GPUs such as a 3090, and is available via the Qwen repository, Hugging Face, GitHub and ModelScope, with a Hugging Face demo.AlibabaのQwen AIチームは、文章からの画像生成と画像編集を1つの流れで行えるオープンウェイトモデルQwen-Image-2.1を公開しました。画像生成部分は70億パラメーターで、Qwen自身のベンチマークでは多くの非公開モデルを上回るとしています。独立したベンチマークでの確認はまだです。透明部分を持つRGBA画像の生成・編集に標準対応し、丸やマスク、塗りつぶしで指定した部分の編集や、最大10枚の参照画像を使った合成ができます。集合写真、バーチャル試着、部屋のデザインなどを用途として挙げています。複数の縦横比でネイティブ2K出力に対応し、3090などの一般向けGPUで動きます。Qwenのリポジトリ、Hugging Face、GitHub、ModelScopeで提供され、Hugging Faceにはデモもあります。
Why it matters何が変わるのか
The model gives developers and businesses access to a capable image generation and editing tool that runs on consumer-grade hardware rather than requiring costly infrastructure. Its research license bars commercial use, so companies wanting to deploy it commercially must apply to Qwen for a separate license.開発者や企業が、高価な専用設備ではなく一般向けの機器で動く画像生成・編集ツールを利用できるようになります。ただし研究用ライセンスでは商用利用が認められておらず、商用導入したい企業はQwenに別のライセンスを申請する必要があります。
What’s nextこれからの動き
Independent benchmarks of Qwen-Image-2.1's claimed performance are still pending.Qwen-Image-2.1がうたう性能について、独立したベンチマークでの確認が待たれます。
A little context · terms in this story理解の手がかり · この記事の用語
- Open weights
- The model’s learned parameters are available to download. The license still determines how you may use or redistribute them; open weights does not necessarily mean open source.AIが学習したパラメーターをダウンロードできること。利用・再配布の条件はライセンスで決まり、オープンソースと同じ意味とは限りません。
- GPU
- A processor designed to perform many calculations in parallel. GPUs are used for graphics and for many AI training and inference workloads.多くの計算を並列に処理する半導体。画像処理のほか、AIの学習や推論にも広く使われます。
- Benchmark
- A test used to compare performance under particular conditions. A higher score on one test does not establish that a product is better for every real task.特定の条件で性能を比較するテスト。あるテストで高得点でも、実際のあらゆる用途で優れているとは限りません。
Read the sources原典を読む
These are the source links saved with this story. Older records do not identify which texts were used in the summary.この記事に保存されている出典です。過去の記事には、どの本文を要約に使用したかの記録がありません。
More in AIAIのほかの記事
SpaceXAI releases Grok 4.7 for coding and knowledge workSpaceXAI、コーディングと知識労働向けにGrok 4.7を公開
3 linked sites関連3サイト
Z.ai open-sources ZCode after covert code-upload backlashZ.aiのZCode、無断アップロード発覚で公開
7 linked sites関連7サイト
Alibaba plans 5T-10T parameter AI model, new chipAlibaba、5兆-10兆パラメータAIモデル計画
7 linked sites関連7サイト