China Telecom AI released Xing4.0-29B-A4B on September 22, 2026, which the model card calls "the first model of this scale trained entirely on the Ascend NPU platform with the MindSpore framework," ru…
Developers guyan364, kurisu6912 and LyricZhao released DeepJIT, a header-only C++20 JIT runtime that gives C++ and Python extension authors one shared interface for compiling, caching, loading and lau…
Z.ai confirmed on August 26, 2026 that its anonymous model 'Ox Alpha' is GLM-5.3-Flash, a 320-billion-parameter mixture-of-experts model with 18 billion active parameters per token, trained entirely o…
BCG's 'The Great Divide' report says the US and China are building increasingly incompatible AI stacks, with the US relying on NVIDIA chips and CUDA, US hyperscaler clouds, and closed models from Open…
Researchers from Huawei have developed proxy models that reduce accelerator requirements by 50%-87.5% and cut per-step NPU-hour cost by up to 33.3x for reproducing failures in large language model rei…
Z.ai launched GLM-5.2, the first open-weight model to top the HuggingFace Open LLM Leaderboard and beat proprietary coding products on SWE-bench Verified with a 72.3% score. Trained on Huawei Ascend h…
Beijing-based Z.ai released GLM-5.2, an open-weight model trained on Huawei Ascend chips without Nvidia hardware, which quickly topped public leaderboards including Design Arena and Artificial Analysi…