AI News 7 min read

Qwen-Image-2.1 on RTX 5090 and 4090: VRAM, INT8, and the New Licence

ai.rs Sep 20, 2026
Qwen-Image-2.1 on RTX 5090 and 4090: VRAM, INT8, and the New Licence illustration

Alibaba's Qwen team released Qwen-Image-2.1 today with open weights on Hugging Face and ModelScope: one model that generates images from text, edits them from up to ten reference pictures, and outputs native transparency, at 2048×2048 out of the box. It is the first open-weight release in the line since December. It is also the first one that is not Apache 2.0, and that second fact will matter to more of our readers than the first.

What it is

The original Qwen-Image from August 2025 was a 20-billion-parameter dual-stream diffusion transformer that made its name on text rendering, especially Chinese. Qwen-Image-2.0 followed in February 2026 as an API-only product; seven months later its weights still have not been published. Version 2.1 goes back to open weights, and it is a smaller, different design:

Qwen-Image (2025) Qwen-Image-2.1
Diffusion transformer 20B, dual-stream MMDiT 7B, 32 single-stream layers
Text and image conditioning Qwen2.5-VL Qwen3-VL 8B
VAE 16-channel RGB 64-channel RGBA, 16× compression
Native resolution 1328×1328 class 2048×2048, seven aspect ratios to 2752×1536
Editing separate Qwen-Image-Edit model same model, up to 10 reference images
Transparency separate Layered model native
Licence Apache 2.0 Qwen Research License

The architectural idea is a single stream with block-causal attention: text tokens attend causally, image tokens attend bidirectionally within their block, and the prompt plus any reference images are encoded once at the first denoising step and reused as a cached prefix for the other 39. That is the same trick as a language model's KV cache, applied to a diffusion model, and it is why an edit with several reference images does not cost several times a plain generation.

The post also folds in two capabilities that used to be separate products: Qwen-Image-Layered, the transparency model from December 2025, is now a prompt away inside the main model, and editing has grown task coverage the earlier line did not advertise, including panoramas from a single selfie, infographics from a product photo and storyboards from a three-view character sheet. All of those are shown as examples, none as measurements.

Two companion models ship with it: prompt rewriters for text-to-image and for editing, both fine-tuned from Qwen3.5-VL 9B, which expand a short prompt into the long descriptive kind the model was trained on and pick an aspect ratio for you. Day-zero support is unusually broad. Diffusers has a pipeline class, ComfyUI has native nodes and two template workflows, vLLM-Omni serves it with FP8 and CUDA graphs, SGLang has a cookbook down to consumer cards, LightX2V has tuned configs for the RTX 5090, and FlagOS publishes prebuilt images for eight non-Nvidia chip platforms.

The licence, read in full

The earlier line, Qwen-Image, Qwen-Image-2512 and Qwen-Image-Edit-2511, is Apache 2.0. Qwen-Image-2.1 is released under the Qwen Research License Agreement, dated today. We read it. The operative clauses:

  • Rights are granted "FOR NON-COMMERCIAL PURPOSES ONLY", and the agreement defines non-commercial as "for research or evaluation purposes only".
  • "You shall not use the Materials for any commercial purpose without obtaining a separate commercial license from us", with an email address for requests.
  • Redistribution and derivatives are allowed, with an attribution notice naming Hangzhou Tongyi Laboratory.

So a designer generating product shots for a client, a shop making banners, an agency building a tool on it: all of that needs a commercial licence that does not yet have a published price. This is the pattern Qwen used for its largest language models and it is new for the image line. If your workflow is built on the Apache-licensed Qwen-Image-2512 or Edit-2511, nothing about those changes; they remain usable for anything. If you were planning to swap them for 2.1, the licence is the first thing to send to whoever signs contracts.

What it needs to run

The README does not say. The vendor's own inference recipes do, and the numbers are worth a table because the headline "7B" undersells them: the text encoder is bigger than the diffusion model.

component BF16 weights
Qwen3-VL 8B text encoder 17.5 GB
DiT, 7B 14.2 GB
RGBA VAE 1.4 GB
total ~33 GB
configuration, 1024×1024 peak VRAM source
all BF16, resident, 40 steps 34.0 GB vLLM-Omni on GB300
all BF16, 50 steps 40.0 GB vLLM-Omni on GB200
text encoder FP8, DiT BF16 27.5–28.1 GB vLLM-Omni on GB300
everything FP8 33.4 GB vLLM-Omni on GB200

Fully resident in BF16, it does not fit a 32 GB RTX 5090, and SGLang's cookbook says so directly: the 5090 and 4090 "exceed single-GPU capacity and require offload". There are two ways out, and both shipped on day one.

Offload the text encoder. It runs once per prompt, so it can live in system RAM and stream through the card layer by layer while the 14.2 GB DiT and the VAE stay resident. That is SGLang's recommended layout for both consumer cards. LightX2V goes further with an FP8 DiT and publishes the only consumer-card timing we have found:

RTX 5090, FP8 DiT, 40 steps, 1024×1024 seconds
text to image 5.93
image edit 7.14

A datacentre GB300 in BF16 does the same image in 3.3 to 4.5 seconds, so the consumer card is not far behind once memory is solved.

Or quantise everything. Comfy-Org's official repackage, published the same day, carries INT8 versions of both big components alongside the BF16 originals, and the file sizes settle the question for the ComfyUI crowd:

component BF16 INT8 (convrot) smaller still
diffusion transformer, 7B 13.3 GiB 6.8 GiB NVFP4 3.9 GiB
text encoder, Qwen3-VL 8B 16.3 GiB 8.7 GiB W4A8 5.9 GiB
VAE 0.6 GiB 0.6 GiB 0.6 GiB
total 30.2 GiB 16.1 GiB 10.4 GiB

At 16.1 GiB the entire pipeline sits on a 5090 with 15 GiB to spare for activations at 2K, and it fits a 24 GB 4090 resident with nothing streamed. The 10.4 GiB combination reaches 16 GB cards. The INT8 prompt rewriters are in the same repository at 8.8 GiB each, though they run before generation and never need to share the card with it. What no one has published is a quality comparison between the INT8 and BF16 outputs, so treat the small files as a fit, not yet as a free lunch.

Every published speed and peak-memory figure is at 1024×1024. Native output is 2K, four times the pixel area, and nobody has published a peak-memory number at that size. That is the measurement we intend to take. A 128 GB GB10 box runs the whole pipeline resident in BF16 with room to spare, which makes this the same capacity-versus-bandwidth trade as every other model we have run this year, only with a 5090 that now costs three times its list price.

The benchmark is Qwen's own

The launch post carries exactly one comparison: a chart of Qwen-Image-2.1 against "other open-source and closed-source models" on Qwen-Image-Bench. That is the evaluation suite Qwen itself published in May 2026, so the vendor is grading its model on its own exam, and the scores appear as a figure rather than a table. The GitHub README and the Hugging Face model card contain no benchmark at all, and no independent evaluation exists six hours after release. No verified comparison against Flux, Seedream, GPT-Image or the 20B predecessor exists yet. The "improved typography, portrait lighting and fine details" in the announcement are the vendor's words, illustrated with the vendor's examples.

What we can say is what the design promises. A 7B transformer with a cached prefix is built for cheap editing rounds, not for winning a quality leaderboard against 20B models; native RGBA and ten-image conditioning are built for product and design workflows; and a research licence is built to be sold. Whether the quality holds up at 7B is exactly what a local test will show, and the text-rendering check will be in Serbian, both scripts, as it was for DeepSeek-V4-Flash.

Bottom line

  • Open weights again, after 2.0 stayed behind an API: 7B DiT, Qwen3-VL 8B encoder, RGBA VAE, native 2K, editing and generation in one model.
  • Not Apache any more. Research and evaluation only; commercial use needs a separate licence with no published price. The Apache-licensed 2512 and Edit-2511 models are unaffected.
  • 33 GB of BF16 weights, 16 GiB in INT8. A 5090 runs BF16 with the text encoder offloaded, at about 6 seconds per 1K image with an FP8 DiT; Comfy-Org's INT8 repackage puts the whole pipeline on a 5090 or a 4090 with nothing streamed. Resident BF16 starts at 48 GB.
  • No benchmarks we could verify. Vendor claims only, until someone measures it.

Frequently Asked Questions

Is Qwen-Image-2.1 free for commercial use? +

No. It is released under the Qwen Research License Agreement dated 20 September 2026, which grants rights for non-commercial purposes only and defines non-commercial as research or evaluation. Commercial use requires a separate licence requested by email from Qwen, with no published price. The earlier Qwen-Image, Qwen-Image-2512 and Qwen-Image-Edit-2511 models remain Apache 2.0 and are unaffected.

How much VRAM does Qwen-Image-2.1 need? +

In BF16 the weights are about 33 GB: a 17.5 GB Qwen3-VL 8B text encoder, a 14.2 GB diffusion transformer and a 1.4 GB VAE. vLLM-Omni measured 34.0 GB peak fully resident at 1024×1024 and 40 steps, so it does not fit a 32 GB RTX 5090 that way. With the text encoder offloaded to system RAM the card holds about 15.6 GB. Comfy-Org's INT8 repackage totals 16.1 GiB for the whole pipeline, which fits a 5090 or a 24 GB 4090 resident, and a W4A8 encoder plus NVFP4 transformer come to 10.4 GiB. No peak figure at native 2K has been published.

How fast is Qwen-Image-2.1 on an RTX 5090? +

LightX2V measured 5.93 seconds for a 1024×1024 text-to-image generation and 7.14 seconds for an image edit, at 40 steps with an FP8 transformer and the text encoder offloaded, as a median of three runs. A datacentre GB300 in BF16 takes 3.3 to 4.5 seconds for the same image. Nobody has published a 2K timing on a consumer card yet.

What changed from Qwen-Image to Qwen-Image-2.1? +

The 2025 Qwen-Image was a 20B dual-stream MMDiT with a separate editing model and a separate layered model for transparency. Version 2.1 is a 7B single-stream transformer with 32 layers and block-causal attention, a Qwen3-VL 8B encoder, a 64-channel RGBA VAE, native 2048×2048 output, editing from up to ten reference images and native transparency, all in one model. A cached prefix means the prompt and reference images are encoded once per generation rather than at every step. Qwen-Image-2.0, released in February 2026, was API-only and never got open weights.

Does Qwen-Image-2.1 work in ComfyUI? +

Yes, natively from day one. Comfy-Org publishes repackaged weights in BF16 and INT8 for the diffusion transformer and the text encoder, plus a BF16 VAE and INT8 versions of the two prompt-rewriter models, with template workflows for text-to-image and image editing. Diffusers, vLLM-Omni, SGLang and LightX2V also support it from release day.

How good is Qwen-Image-2.1 compared to Flux or GPT-Image? +

No verified answer exists yet. The launch post's only comparison is a chart on Qwen-Image-Bench, Qwen's own evaluation suite from May 2026, against unnamed open and closed models. The GitHub README and Hugging Face model card contain no benchmark at all, and no independent evaluation exists on release day. The announcement's claims of improved typography, portrait lighting and fine detail are the vendor's words. A local test on an RTX 5090, including Serbian text rendering in both scripts, is the follow-up we intend to publish.

What does this mean for your business?

New models drop every month. The real question is whether the underlying capability fits your business. Find out in 2 minutes.

Take the AI Readiness Check
Share: Post Share

Read next