Qwen-Image Pricing Explained (2026): Free App & API Costs

Qwen-Image pricing splits into two very different stories. The newest model, Qwen-Image-3.0, is free at chat.qwen.ai but has no published API rate yet. The older Qwen-Image 2.x line is open-weight and already priced on third-party hosts from under a cent per image. If you need production access today, you're buying 2.x, not 3.0.

Route	Price	What you get
chat.qwen.ai (Qwen-Image-3.0)	Free	Full 3.0 model in browser; no API, rate-limited by the app
Self-host Qwen-Image 2.x (open weights)	Free (your GPU)	Apache-2.0 weights; you pay only compute
fal.ai / Replicate — Qwen-Image 2.x (t2i)	from ~$0.0058/image (~$0.02/MP)	Pay-per-image generation API
fal.ai — Qwen Image 2.0 edit	~$0.035/image	Style transfer, object insert/remove, text overlay
Alibaba Cloud Model Studio	~$0.10/image baseline	Official managed API (image gen)
Qwen-Image-3.0 API	Unpublished (as of Jul 2026)	Not yet on the official price list

How Qwen-Image pricing works

There isn't one Qwen-Image price — there are two models with two completely different distribution setups. Alibaba's Qwen (Tongyi Lab) team ships its newest model straight into the chat app first, closed, and only opens an API and weights later once it's stable. That's the pattern with 3.0 right now.

The older 2.x generation already went through that cycle. Its weights are public under Apache-2.0, so third-party inference hosts like fal.ai and Replicate picked it up and set their own per-image rates. Alibaba also sells it through its own Cloud Model Studio API at a higher, more enterprise-flavored price. Which route you use depends entirely on whether you need 3.0's quality or can live with 2.x's cost.

Qwen-Image-3.0: free but closed

Qwen-Image-3.0 launched on July 21, 2026, and as of this writing it's free to use at chat.qwen.ai with no per-image charge. That's the only official way to run it — there's no API, and Alibaba hasn't released downloadable weights, so self-hosting isn't an option either.

The app itself rate-limits usage rather than billing you, so heavy users will hit soft caps before they hit a paywall. For a full look at what the model actually produces, see our Qwen-Image-3.0 review, and for how it stacks up against Google's model, check Qwen-Image-3.0 vs Nano Banana 2.

The practical catch: nothing here is buildable into a product. No API means no automation, no batch jobs, and no way to embed 3.0 in your own app until Alibaba publishes rate-card pricing and, ideally, adds it to Model Studio.

Qwen-Image 2.x API costs

If you need a working API today, you're using Qwen-Image 2.x, and it's cheap. On fal.ai and Replicate, the entry tier starts around $0.0058 per image, with text-to-image pricing that generally works out to roughly $0.02 per megapixel.

Editing costs a bit more. The unified Qwen Image 2.0 edit model runs about $0.035 per image on fal.ai, while the original Qwen Image Edit model is priced closer to $0.03 per megapixel. Some hosts charge up to $0.075 per image for higher-resolution generation or edit jobs, so it's worth checking the specific host's tier before committing to a workload.

Alibaba's own Cloud Model Studio API prices image generation at roughly $0.10 per image as a baseline — noticeably above the third-party hosts. You'd choose Model Studio for enterprise SLAs and Alibaba Cloud integration, not for the lowest per-image cost.

Self-hosting the open weights

Qwen-Image 2.x ships under an Apache-2.0 license, so you can download the weights and run inference on your own hardware for free — you only pay for GPU time. That makes it the cheapest option at real volume, assuming you already have (or can rent) the compute and the ops capacity to serve it reliably.

This only applies to 2.x. Qwen-Image-3.0 has no public weights, so there's nothing to self-host yet. If Alibaba follows its usual pattern, an open release could follow the closed chat launch, but there's no confirmed timeline as of July 2026.

Is it worth it?

For casual or one-off use, yes — chat.qwen.ai gives you 3.0 quality at zero cost, and it's the best free option if you don't need automation. For anything that has to run programmatically, Qwen-Image 2.x is the worthwhile pick: sub-cent-to-a-few-cents per image on fal.ai or Replicate is competitive with almost every other production image API on the market, and self-hosting pushes the marginal cost even lower.

What isn't worth doing yet is waiting on a Qwen-Image-3.0 API. There's no confirmed release date for pricing, so budgeting around it is speculative. Compare 2.x's real numbers against the field in our best AI image generators roundup, and if Google's model is also on your shortlist, see the current rates in Nano Banana 2 pricing.