Realtime AI News
Qwen posts Qwen-Image-2.1-PE in two variants: text-to-image and image editing
On September 20, Qwen added two entries to its official Hugging Face organization: Qwen-Image-2.1-PE-T2I for text-to-image generation and Qwen-Image-2.1-PE-I2I for image editing. Both carry prompt-rewriting and qwen3_5 tags under a license:other license, and both showed zero downloads at capture time.

On September 20, the Qwen team added two entries to its official Hugging Face organization: Qwen/Qwen-Image-2.1-PE-T2I and Qwen/Qwen-Image-2.1-PE-I2I. Both are listed as official model registry updates rather than third-party mirrors.
The names make the split clear. The T2I repository is registered for a text-to-image pipeline, while the I2I repository is tagged for image editing, so the same model line now covers both generating an image from a prompt and modifying an image that already exists.
The repository tags add detail. Both entries carry safetensors, qwen3_5 and qwen tags, and both are tagged prompt-rewriting, which points to a step that rewrites a user's wording into an instruction the image model handles better.

Both entries are licensed under license:other with region:us, so the terms are not a standard permissive open-source license. Teams planning commercial use will need to read the model card terms before adopting either checkpoint.
Adoption signals are still minimal. At capture time both repositories showed zero downloads, and their like counts were 1 and 0 respectively, meaning the listings are brand new and no public third-party testing exists yet.
Shipping generation and editing under one banner makes it easier to build consistent workflows that rewrite a prompt, render an image, and then edit specific regions. For product teams that generate at volume and iterate, that pairing may matter more than a single benchmark win.
Important details are still missing. The candidate information does not include parameter counts, training data, benchmark results, or whether the models will also appear on Qwen's hosted services and domestic model platforms, and the meaning of PE in the name still needs confirmation from the official model card.
Three things are worth watching: whether the model cards fill in architecture and evaluation details, whether the license permits commercial use, and whether these variants open up on Qwen's cloud endpoints and domestic platforms at the same time.
Sources
Why it matters
Covering generation and editing under one model line shows the competition shifting from raw image quality to instruction-following edits. The license:other terms also mean enterprises must check commercial boundaries before adopting either checkpoint.
Nearby Updates
All09/20, 19:48
Alibaba's Qwen3.8-Omni-Flash billed as one model that hears, sees and acts
The Eastern Herald describes Alibaba's Qwen3.8-Omni-Flash as a single model that hears, sees and acts. The framing points to a unified system that combines audio and visual understanding with the ability to take action, instead of stitching separate models together.
09/20, 20:22
China Telecom open-sources Xing4.0-29B-A4B, a 29B MoE coding agent that runs on one RTX 3090
China Telecom AI has open-sourced Xing4.0-29B-A4B, a lightweight coding-agent model with 29 billion total parameters and roughly 4 billion active per pass, small enough after 4-bit quantization to run on a single RTX 3090. It was trained entirely on Huawei Ascend 910C hardware with the domestic MindSpore stack, and is already listed on GitHub, Hugging Face, Gitee and ModelScope.
09/20, 21:02
Tempus secures up to $9.5M from ARPA-H for an autonomous AI agent in heart failure care
Tempus has secured up to $9.5 million from ARPA-H to develop an autonomous AI agent for heart failure care, according to Pulse 2.0. The award pushes agent-form software into long-term chronic care, where authorisation limits and accountability have to be settled before deployment.
09/20, 18:50
Another Chinese MoE flagship lands third in the open-source ranks, with pricing aimed at DeepSeek-V4-Pro
Zhidongxi reports that another Chinese MoE flagship model has arrived, placing third among open-source models worldwide while pricing itself against DeepSeek-V4-Pro. The launch raises the stakes on both the leaderboard and the price competition among Chinese open-weight releases.