10 results on this page
AntLing has released six official Ling-3.0 base checkpoints spanning two model sizes and three training stages. The MIT-licensed, ungated repositories are intended for continued pretraining, fine-tuning, and research rather than direct chat use.
Tencent's Hunyuan Hy4 has reportedly appeared in the model selection list of the Yuanbao app under an expert-level label and with tool-use capabilities. It is positioned above Hy3 and alongside DeepSeek, following Tencent's recent statement that a larger-parameter Hy4 would launch soon with improved performance and multimodal abilities.
An independent builder pretrained a 1.02-billion-parameter Kimi-K3 replica on 5 billion decontaminated tokens for roughly $250. The model achieved 33.4% on HellaSwag, exceeding the reported 28% score of GPT-2 124M despite receiving no instruction tuning.
The paper proposes Transition Complexity Profile (TCP), a reproducible framework for measuring how difficult game-world transition prediction is at a specified interface. It evaluates branching, interaction-driven uncertainty, opponent influence, and temporal or spatial dependencies to improve comparisons across game-modeling and reinforcement-learning benchmarks.
NVIDIA FLARE enables organizations to coordinate training for vision-language models while keeping sensitive image and text data at local institutions. The approach supports multimodal use cases such as visual question answering, captioning, and image-text reasoning without centralizing raw records.
Unsloth released new Qwen3.8-27B GGUF quantizations claiming 10% higher accuracy at the same file size. The release also includes 1-bit quantizations that retain 77% accuracy and can run with 8GB of RAM.
Loudoun County, Virginia, has used more than 250 data centers to generate substantial local wealth and reduce property taxes. Its success is attracting attention from other jurisdictions despite growing national opposition to data center expansion.
Aurora-80K is a tiny language model with exactly 80,000 parameters and a factorized 4,096-token vocabulary. Its reported results include 3.2902 Wikitext-2 BPB, 52.31% BLiMP accuracy, and 26.05% on ARC-Easy.
Alibaba will sell all shares of Lingxi Games to Chixin Capital for 10.1 billion yuan, exiting self-developed game content. The deal reflects Alibaba’s shift from gaming toward AI, where its Qwen models and consumer products are becoming strategic priorities.
A Reddit user reports that Qwen3.8-27B performs substantially worse than Qwen3.6 on obscure trivia and offline knowledge tests. The model may still be strong for coding and tool-assisted workflows, but users relying on its internal weights for broad factual recall should be cautious.