Hy Image3.5 preview is now available in ComfyUI. Professional-grade image generation, +30% win rate in human eval vs Hy Image3.0 → Text to image and Image to image in one model, up to 2K → Multilingual text, symbols, and small print that render correctly → Cinematic, comic, commercial photography, and illustration styles → Identity and product features that hold through scene, outfit, and style changes

Hy Image3.5 preview is live on Miora. FREE through October 7. Campaign visuals. Product scenes. Storyboards. Game concepts. Up to 2K. Text-to-image and image-to-image. Try it:

Hy Image3.5 preview is live. 🚀 Professional-grade image generation, +30% win rate in human eval vs Hy Image3.0 Both Text to image & Image to image available. Up to 2K. Better Consistency. API: Priced for everyone. $0.024 per image on Tencent Cloud API. We only charge for what we generate — your reference images are free. Two weeks free (Only in OnSolo and Miora): Miora OnSolo Try it and tell us where it breaks.

How Tencent Packed a 770B-Parameter Model into 214 GiB Shrinking Hy4 preview's weights from roughly 1.5TB to 214 GiB is one challenge. Preserving useful capabilities and practical inference speed is another. How did @TencentHunyuan tackle both? Zhihu contributor yghstill, a member of Tencent Hunyuan's quantization team, explains the engineering behind it. The parameter count remains 770B; the compression changes how those weights are represented. Four weights, five bits Sherry is the quantization algorithm, STQ1_0 the storage format, and MIX-STQ1_0 the mixed-precision allocation scheme. Each group of four weights takes values from {-d, 0, +d}, with exactly one zero. Four zero positions multiplied by eight sign combinations gives 32 possible patterns, requiring five bits. That is 1.25 bits per weight for the codes alone. Including a shared FP16 scale for every 256 weights brings STQ1_0 to 1.3125 bits per weight. The complete mixed-precision model averages about 2.38 bits per weight. Allocate precision to specific weights, not whole layers Hy4 preview has 77 MoE layers, each with 256 routed experts. The team concentrates aggressive compression on expert weights while selectively protecting other components. For the experts' gate/up projections, MIX-STQ1_0 uses IQ2_XXS on 48 sensitive layers and STQ1_0 on 29 less sensitive layers. The author reports that mixing lower and higher precision produces less error at the same average bit budget than uniformly choosing the intermediate IQ1_M format. Layer sensitivity needs more than diagonal statistics The author describes using the full Hessian, H = XXᵀ, to measure quantization sensitivity. Its off-diagonal terms capture correlations that diagonal-only imatrix scoring misses. In the team's comparison, the two sensitivity rankings had a Spearman correlation of -0.115. The chosen layers did not follow a simple “deeper means more important” rule: precision was allocated greedily by error reduction per additional byte. Fit the scale and choose the zero together This is post-training quantization, without retraining. The encoder alternates between two decisions: fitting d with weighted least squares and choosing the zero position using imatrix-weighted error. Zeroing the smallest-magnitude weight is not always the best choice. What matters is the additional weighted error introduced by making it zero. Across 1,200 rows of real expert weights, three alternating rounds reduced weighted reconstruction error by roughly 90% compared with the original ternary encoder. This measures local weight reconstruction, not end-to-end model accuracy. Compression must survive the runtime The team implemented STQ1_0 CUDA kernels in a patched llama.cpp build. In its operator comparison, STQ1_0 ran roughly as fast as IQ1_M despite the lower bit width. The author reports nearly unchanged MRCR retrieval performance and a small decline in math. Against UD-IQ1_M at a similar bit budget, the mixed-precision model led across the reported evaluations, including a gain of more than five points on MRCR. The result comes from combining compact encoding, calibrated quantization, selective precision and usable inference kernels.

This morning I visited Tencent HQ in Beijing! A marvelous office, from the architecture to the amazing showroom where I've really understood values and services offered by Tencent. Thanks to the @TencentHunyuan team in Beijing for the amazing lunch talking about hybrid AI scenarios and more! Hy4 Preview is just the tip of the iceberg, there is much more to come! @TencentAI_news Can't wait to see and try what you'll deliver in the near future 🚀