FULL STORY

Tencent Hunyuan HyImage3.5: From Launch to Hands-On

Tencent Hunyuan launched HyImage3.5 preview on Sept 22, touting 2K image generation at just $0.02, followed by ComfyUI availability and hands-on reviews highlighting its text rendering and editing control.

2026-09-22 ~ 2026-09-23 · 2 episodes · 18 posts

Episode 1 · Tencent Hunyuan Image 3.5 Preview Rolls Out with 30% Win-Rate Gain (2026-09-22, 10 posts)

On September 22, Tencent Hunyuan officially released its next-gen image generation model HyImage3.5 preview, touting professional-grade generation at aggressive pricing, and launched it exclusively on third-party platform OnSolo. It's a routine generational upgrade for the Hunyuan image series, but drew attention with 2K-resolution output and industry-low pricing.

Confirmed

  • Officially announced: HyImage3.5 preview supports text-to-image and image-to-image, up to 5 reference images per generation, multiple aspect ratios, up to 2K resolution, and multi-turn conversational editing.
  • In a GSB blind test with hundreds of internal designers, overall performance improved about 30% over HyImage3.0 (30% higher win rate in human evaluation).
  • API pricing is $0.024/image (about RMB 0.15 per image), which Zhidx reported as the lowest in the industry.
  • The model launched exclusively on OnSolo, supporting 5 reference image inputs and 2K output at 1 credit per generation; during a two-week promotional period, OnSolo members get unlimited free generations.

Why it matters

  • Ultra-low pricing paired with 2K output directly disrupts image generation market pricing—Zhidx called it the industry's lowest price.
  • The exclusive OnSolo launch with free member generation is a textbook case of community model distribution. Sharer @Divpradeep and tester @LearnWithBishal both reported solid generation quality, fast speed, and competitive pricing.

Episode 2 · Hands-On: Tencent Hunyuan Image 3.5 Shines in Text Rendering and Editing Control (2026-09-23, 8 posts)

On September 23, creator HeyAmit posted a thread of hands-on tests with the Tencent Hunyuan Image 3.5 (Hy Image 3.5) preview. His overall take: the model's value goes beyond image quality — it lies in fine-grained control over the generation process and in embedding image generation into a more complete workflow.

Confirmed

  • Supports both text-to-image and image-to-image modes, with up to 5 reference images usable as visual context rather than relying on a single prompt.
  • For text generation, it can render text across multiple languages, font sizes, and layouts, and understands how typography relates to the composition — the creator found it especially well suited for posters, ads, presentations, and UI concept drafts.
  • For image editing, in tasks like scene replacement, style adjustment, and image transformation, the design goal is to preserve key features of the reference image while applying the changes requested by the user; in testing, changing scenes and adjusting styles still preserved the reference image's characteristics.
  • Workflow extensions: the creator listed use cases ranging from product shots to ad creatives, ideas to posters, game concepts to UI, and reference images to new visuals, arguing that the model's value goes beyond one-off "generate me an image" requests.

Why it matters

Text rendering has long been a pain point in AI image generation, while multi-reference input and editing with feature preservation determine whether a model can fit into real production pipelines. If performance holds up in practice, Hy Image 3.5 could become a practical tool for posters, advertising, presentations, and UI design. It is currently a preview, so the final release remains to be seen.