AI Lab
Official Tencent Hunyuan announcements — model releases, product launches and research, each one summarized with every source covering it, by AIBriefs. RSS
AuK unifies speech generation and editing via natural-language instructions and audio context, built on a multimodal language model, joint VAE, and hybrid rectified-flow Transformer. The AuK-Flash variant does fast 4-step inference via distillation.
Hy4 preview has 770B total parameters, 49B active, and a context window exceeding 1M tokens, up from Hy3's 295B/21B and 256K context in July. It scored 2.99/4.00 in Tencent's blind eval of 163 experts across 203 engineering tasks, ahead of GLM-5.3 (2.92) and Kimi K3 (2.94).
Tencent released WeMM-Embedding-2B on HuggingFace, a 2B-parameter embedding model. It has gained 68 likes and over 5,400 downloads.
Tencent's EVIE-8B achieves 66.75 nDCG@10 on ViDoRe V3, a SOTA result. EVIE-4.5B is built on Qwen3.5-4B with 4.54B parameters and native 128-dimensional multi-vector embeddings.
The 27B model observes live screenshots, reasons over the visible state, and outputs structured keyboard and mouse actions for long-horizon native desktop interaction across applications and operating systems.