Access guide

How to use Hy Image 3.5 Preview

Hy Image 3.5 Preview is a hosted model, so using it means choosing an access point. Tencent offers it in its own products and through an API, and a few partners have added it. Here is each route, who it suits and what it costs.

Reviewed October 4, 2026 · Independent guide, not affiliated with Tencent

Generate an image free Jump to FAQ

Where Hy Image 3.5 Preview is available

Route	Best for	Cost and limits
Tencent Cloud TokenHub API	Developers who want the full limits	$0.024 per image per Tencent's launch post; up to 20 references in the API guide
Miora	Designers who want a web tool	Two-week free trial announced at launch
OnSolo	Video and content workflows	At launch: two weeks free for existing members, 110 free images for new sign-ups, per press coverage
Yuanbao, WorkRally, WorkBuddy, ima	People already using Tencent's apps	Integrated by Tencent; terms are set in each product
ComfyUI partner nodes	Node-based workflows	Billed through a Comfy account; up to 5 references
GMI Cloud	A non-Tencent API and playground	Listed at $0.024 per image under `hy-image-v3.5-preview`

Tencent names Miora and OnSolo, together with the Tencent Cloud API, as the routes for international users. The other products belong to Tencent's wider app ecosystem, and their availability varies by region.

Can you use Hy Image 3.5 for free?

For a limited time, in some places. Tencent paired the September 22, 2026 launch with a two-week free window in Miora and OnSolo, which coverage reports as closing on October 7, 2026. Free access after that date depends on each product's own plan, so check inside the product rather than relying on a launch-week article. The API has no free tier in Tencent's announcement: it is $0.024 per output image.

This workspace gives every new account one free image: a single Hy Image 3.5 Preview generation (5 credits). After that, each model shows its own credit quote before you submit.

Using it through the API

Create an API key in the Tencent Cloud TokenHub console.
Send a POST request to the v35-generation endpoint with model set to hy-image-v3.5-preview and your brief inside a messages array.
Read the image URL from choices[0].delta.image.url and download it within 12 hours.
For another round, append the returned assembled_history to your messages and add one new instruction.

Full Hy Image 3.5 API guide

Using it in ComfyUI

Update ComfyUI, sign in to your Comfy account, and open the Tencent HY Image 3.5 Preview template for text-to-image or image edit. The nodes run on Tencent's servers, so there is nothing to download.

Hy Image 3.5 in ComfyUI, in detail

Working in rounds

What separates this model from a one-shot generator is that a result can be revised across turns. The way requests are structured suggests a simple habit, whichever route you use.

First turn: describe the subject, composition and light, and attach the references that matter. Say what each reference is for.
Each later turn: ask for one change and name what must stay. "Keep the bottle and the label, change the background to slate" is easier to judge than five edits at once.
If text has to appear in the image, write the exact wording in quotation marks and check it at full size. Tencent lists text rendering as an improved area; an independent hands-on found small text and numbers less reliable than headlines.
Save the version you want before continuing. API result links expire after 12 hours.

Preview limits worth knowing

Resolution: Tencent's launch post says up to 2K. The API guide additionally accepts an explicit size up to 4096x4096.
References: five in Tencent's products and the ComfyUI node, up to 20 in the API guide.
Availability: it is a preview. Rate limits apply, and features or pricing can change without a new version number.
Weights: none are public, so there is no local or offline option.

What you can do on this site today

hyimage35.pro is an independent workspace, not affiliated with Tencent. Hy Image 3.5 Preview is the default model here, with 2K or 4K output and a separate edit mode that takes up to 20 reference images. Hunyuan Image 3.0 and its Instruct routes remain available under their own names.

Write the brief: subject, composition and light in one clear paragraph.
Keep Hy Image 3.5 Preview selected (5 credits per image), or switch to Hy Image 3.5 Preview · Edit and attach the references you want to work from.
Check the credit quote, generate, and download the result from the workspace or your account history.

Hy Image 3.5 vs Hunyuan Image 3.0 Pricing and credits

How to use Hy Image 3.5 FAQ

Is Hy Image 3.5 free to use?

Partly and temporarily. Tencent announced a two-week free window in Miora and OnSolo starting September 22, 2026. The Tencent Cloud API is paid at $0.024 per output image. Check each product for its current terms.

Can I use Hy Image 3.5 without a Tencent Cloud account?

Yes. Tencent's consumer and design products such as Miora and OnSolo include it, ComfyUI offers it through partner nodes with a Comfy account, and GMI Cloud lists it in its own inference engine.

Does Hy Image 3.5 support image-to-image?

Yes. Tencent released it with both text-to-image and image-to-image, and the model edits from reference images plus an instruction. The API guide documents up to 20 references; product integrations describe five.

Can I use Hy Image 3.5 on hyimage35.pro?

Yes. Hy Image 3.5 Preview is the default model here, for text-to-image and for reference editing with up to 20 images, at 2K or 4K. Each image costs 5 credits, and every new account gets one free image.

Can I run Hy Image 3.5 offline?

No. Tencent has not released weights for 3.5. The open-weight model in this family is HunyuanImage 3.0.