News about foundation models, LLMs, multimodal models, benchmarks, and model releases from OpenAI, Anthropic, Google, Meta, Mistral, and others.
715 articles
FTC Chairman Andrew Ferguson says competition among large language model developers is 'pretty ferocious,' but warns that antitrust risks could emerge further upstream in the AI supply chain. Source: broadbandbreakfast.com
A new analysis looks at how generative AI and large language models are rapidly being adopted in clinical care, while highlighting the safety and security concerns this integration raises. Source: nature.com
Researchers used a large language model to automatically generate compilers for shuttling operations in trapped-ion quantum computers, cutting the required timesteps by up to 76%. Source: quantumzeitgeist.com
While open-weight language models keep growing in size, a parallel trend explores extreme compression—using 1-bit models to squeeze capabilities comparable to 27-billion-parameter models onto everyday devices. Source: intelligentliving.co
The Army's Pittsburgh-based AI center is partnering with Washington-based EdgeRunner AI to build a language model that can run locally, cutting reliance on the cloud and reducing cybersecurity exposure. Source: bizjournals.com
KT's new NPU LLM Station pairs Rebellions' ATOM-MAX inference chip with the Mi:dm K 2.5 Pro language model in a single on-premises server, marking Korea's first fully domestic sovereign AI appliance. Source: techtimes.com
According to three sources familiar with the matter, Apple has trained a dedicated large language model tailored to the Chinese market, working with Alibaba for support. The move comes as Apple prepares to bring its Apple Intelligence features to Chinese users. Source: hungarianconservative.com
Baidu CEO Robin Li pledged on Tuesday to restore the company's Ernie large language model to the cutting edge of AI development. The promise follows a disappointing revenue report that sent Baidu's shares tumbling. Source: kfgo.com
EdgeRunner is partnering with the U.S. Army Artificial Intelligence Integration Center to build a large language model tailored for Army needs. The project focuses on open-weight models that can run locally, offering better performance, efficiency, and control for sensitive national security applications. Source: 01net.it
Speaking in Alexandria, Virginia, a DARPA official said that despite the hype around AI tools like large language models, they currently offer limited practical value for military missions on their own. The official emphasized that LLMs need to be paired with other capabilities to be truly useful in defense applications. Source: nationaldefensemagazine.org
Researchers built a biomedical foundation model that combines vision and language pretraining with explicit medical concepts, aiming to make AI diagnoses more interpretable for doctors. The approach addresses a key limitation of current multimodal medical AI systems, which often lack transparency. Source: nature.com
MIT researchers compared brain connectivity patterns among people who wrote essays using ChatGPT, a search engine, or relying solely on their own thinking. The study looks at how reliance on AI tools may influence cognitive engagement during writing tasks. Source: britannica.com
Researchers at Indiana University used the REALLMS platform to run faster and more reproducible experiments with open-source large language models. The infrastructure has helped accelerate their work on making AI systems more transparent and understandable. Source: news.iu.edu
ByteDance Seed and Tsinghua AIR introduced CUDA Agent, a large-scale reinforcement learning system that trains an LLM to automatically generate CUDA GPU kernels. The project aims to speed up and automate low-level GPU code optimization typically done by human engineers. Source: marktechpost.com
A new system built on a medical large language model was designed to train and evaluate how well medical students conduct patient history interviews. Early results indicate the tool improves students' performance in this core clinical skill. Source: cureus.com
Citing three sources, Reuters reports that Apple developed a large language model specifically for the Chinese market in cooperation with Alibaba. The move reflects Apple's push to better tailor AI features to local regulatory and market conditions. Source: forklog.com
Apple is said to be training its own large language model for the Chinese market instead of relying on a third-party model to power AI features there. Alibaba is reportedly supporting the effort as part of a broader partnership. Source: ilounge.com
In December 2025, Starcloud used a data-centre-grade NVIDIA H100 GPU orbiting roughly 500 km above Earth to train NanoGPT on a Shakespeare text dataset. It was a small-scale but genuine demonstration, separate from the satellite's preloaded software. Source: scienceblog.com
A cascade architecture filters requests before they hit the model, avoiding the need to treat every query probabilistically. The approach promises up to 6x lower inference costs for retrieval-augmented systems. Source: venturebeat.com
According to reports, Apple has trained a large language model tailored to the Chinese market with Alibaba's assistance, deepening ties tied to its Qwen partnership. This marks a departure from Apple's earlier approach of relying on third-party models for the region. Source: digitimes.com