Tasks and inputs
Hy Image 3.5 Preview is listed for Image-text-to-image, Image-to-image, Text-to-image. It accepts text, image and produces image. Use the publisher’s input format and preprocessing for your chosen deployment.
SyncDev model directory · Profile updated Nov 17, 2026
Hy Image 3.5 Preview is Tencent's hosted image generation and multi-turn editing model for design work, with reference images and output up to 4K.
The specification score summarizes recorded features and limits. It is not a benchmark result.
Pricing conditions: Reported about USD 0.024 per image (vs USD 0.032 for Hy Image 3.0); a registry listing shows an image-output figure of 1.6 with unclear units. Indicative; confirm on your own account.
Recorded modality support shows what this model can accept and produce. An undocumented modality is shown as unknown rather than assumed to be unsupported.
Accepts
Produces
Ad layouts, UI mock-ups, product shots. Tencent aimed this preview at design jobs where the first image is never the final one, and where the conversation with the model matters as much as the prompt. The model keeps editing context across turns, so "make the headline bigger, now swap the background" works as a dialogue rather than three separate jobs.
It was announced on 22 September 2026. Before that it ran inside Tencent's Yuanbao and ima apps for about two months.
Output is offered in 1K, 2K and 4K tiers, plus custom sizes in multiples of 16 up to 4096 × 4096. Aspect ratios run from 21:9 to 9:21. Here is the catch: anything above 2K is rendered at 2K and then upscaled. Tencent's own positioning is 2K-native, with 4K reached through a specific request path. If you need true 4K detail for print, inspect crops closely.
| Point | Version A | Version B |
|---|---|---|
| Reference images per request | 5 | Up to 20 |
| Price | About $0.024 per image | One listing shows an image-output figure of 1.6 with unclear units |
The five-image limit comes from the workflow tool integration, which addresses references as @image1 to @image5. The higher figure comes from an API reseller and may reflect a different endpoint. Test with six references to find out which applies to your route. As for price, $0.024 per image is 25% below Hy Image 3.0's $0.032; treat it as indicative and confirm on your account.
Tencent ran a blind test with professional designers and reports a gain of about 30% over Hy Image 3.0. It also claims parity with or a slight lead over several competing commercial image models. These are vendor-run results, not independent benchmarks.
This version is hosted only. No weights are published. Access has been through partner nodes in a popular node-based workflow tool, which require an account with that tool's cloud service, and through API resellers. If you need a self-hosted model from the same family, the earlier open-weight Hunyuan Image 3.0 is the option.
The model ID reported for API use is hy-image-v3.5-preview.
Brief it like a junior. Supply a brand logo and a product photo, ask for a square social ad, then make four rounds of edits: copy change, colour change, layout shift, background swap. Check whether earlier edits survive later ones. That is where multi-turn models usually break.
Compare against other image-to-image and text-to-image options. More from the lab is on the Tencent provider page.
Compare the model on the work you need to complete.
Hy Image 3.5 Preview is listed for Image-text-to-image, Image-to-image, Text-to-image. It accepts text, image and produces image. Use the publisher’s input format and preprocessing for your chosen deployment.
Choose the output size and quality your design needs. Include drafts and revisions when budgeting an accepted image.
Use the model identifier and account requirements for the selected API route. Confirm its request limits and output format with one complete task before connecting a larger workflow.
These records share a directory category. Check the source card, output type and deployment requirements before treating a model as a replacement.
recraft
Recraft V4.1 Flash V2 is Recraft's fastest, cheapest image model: 1K raster output in about 1.1 s for $0.007, quality close to Recraft V4.1.
MiniMax
MiniMax H3 Image is an open-weight 33.1B image generation and editing model producing native ~16MP output in six steps, with up to 16 reference images.
MiniMax
MiniMax H3 Image Highspeed is the faster hosted tier of MiniMax's H3 Image generator and editor, at double the image-token price of the standard tier.
OpenAI
GPT-Image-2.6 Flare is OpenAI's fast image generation and editing model for interactive iteration, with SVG, animated GIF and markup-guided edits.
OpenAI
GPT-Image-2.6 Sunburst is OpenAI's detail-first image generation and editing model, adding SVG, animated GIF output and markup-guided edits.
krea
Krea 3 is Krea's in-house image model for text-to-image and instruction-based edits at 1K or native 2K, priced from $0.03 per image.