Tencent has released the Hy-MT2 family of "fast-thinking" multilingual translation models, including 1.8B, 7B, and 30B-A3B (MoE) sizes, along with the IFMTBench benchmark for evaluating instruction-following capabilities.
Tencent has released the UI-Mate-democua-27B checkpoint, a demonstration-guided version of its UI-Mate GUI agent that can adapt recorded workflows to new tasks. The model observes live screenshots and reasons over visible states to produce structured actions, using demonstrations as guidance rather than fixed scripts.
Tencent has released UI-Mate-27B, an open-weight foundation model designed for long-horizon computer use across applications and operating systems. The 27-billion parameter model observes live screenshots to reason over visible states and produces structured keyboard and mouse actions for native desktop interaction.
Tencent HY Frontier Multimodal Agent Team has released UI-Mate-27B, an open-weight foundation GUI agent designed for long-horizon tasks across applications and operating systems. The model observes live screenshots to reason over visible states and produces structured keyboard and mouse actions for native desktop interaction.
Tencent HY Frontier Multimodal Agent Team has released UI-Mate-9B, an open-weight foundation GUI agent designed for long-horizon tasks across applications and operating systems. The model observes live screenshots to reason over visible states and produces structured keyboard and mouse actions for native desktop interaction.
Tencent has released EVIE-Preview-4.5B, a state-of-the-art multilingual Visual Document Retrieval (VDR) model built upon Qwen3.5-4B. The model employs ColBERT-style late interaction with native 128-dimensional multi-vector token embeddings to achieve top-tier performance across ViDoRe benchmarks.
Researchers have developed MiLMMT-46-v1.0, a multilingual machine translation model that applies reference-free post-training to open large language models. The team uses Group Relative Policy Optimization (GRPO) with a reward based on two reference-free quality estimation models, gated by language identification.
Researchers have released MiLMMT-46-v1.0, an open large language model for multilingual machine translation that utilizes reference-free post-training. Starting from the supervised-finetuned MiLMMT-46-v0.1, the team applied Group Relative Policy Optimization (GRPO) with a reward based on two reference-free quality estimation models gated by language identification.
Tencent has open-sourced AngelSpec, a torch-native training framework that supports both autoregressive multi-token prediction (MTP) and block-parallel DFlash-family drafters for Hy3 models. The system treats workload heterogeneity as a primary design constraint, allowing it to specialize structure, training data, and verification depth based on specific use cases.