🧠 Large Language Models

News about foundation models, LLMs, multimodal models, benchmarks, and model releases from OpenAI, Anthropic, Google, Meta, Mistral, and others.

1,576 articles

AI Agents Are Booming — But Research Warns They Leak Private Data

A professor at Wake Forest University has identified serious security flaws in large language models used to power AI agents. The findings suggest these systems can inadvertently expose sensitive personal information. Source: digitalinformationworld.com

Brief Chatbot Chat Outperformed Fact Sheets in Curbing Conspiracy Beliefs, Studies Find

Two experiments showed that a short conversation with an AI chatbot was more effective than a written fact sheet at reducing belief in conspiracy theories. Researchers found this worked even shortly after a crisis, when solid evidence was still scarce. Source: the-decoder.com

AI Diagnoses Rheumatic Diseases Faster Than Doctors, But Not More Accurately

Prof. Valmed, an LLM-based diagnostic tool approved by European regulators, was tested against human physicians in diagnosing rheumatologic conditions. The AI delivered results faster, but its accuracy did not surpass that of human doctors. Source: medpagetoday.com

Saudi Humain model draws attention as Chinese AI tech goes global

Reports about Saudi Arabia's Humain using MiniMax's Chinese AI model as the base for its Arabic LLM have sparked wider discussion about China's growing influence in exporting AI technology. Observers note this reflects a broader trend of Chinese models being licensed or adapted internationally rather than just used via APIs. Source: kucoin.com

What we know about OpenAI's rumored next-gen model 'GPT 6 Astra'

Speculation is growing around a supposed new frontier model from OpenAI, referred to as GPT 6 Astra, covering its potential capabilities, pricing, security risks, and impact on ChatGPT. The article outlines what developers and businesses should consider as they wait for official details. Source: blockchain-council.org

Saudi company builds Arabic LLM on top of Chinese AI model

Saudi state-backed AI firm Humain has released an Arabic large language model built on technology from China's MiniMax. The move highlights how Chinese AI models are increasingly being adopted abroad as a foundation for localized products, not just through APIs or finished apps. Source: globaltimes.cn

Humain's Arabic model M3 confirmed to be based on MiniMax's technology

Saudi Arabia's state-owned AI company Humain officially confirmed that its new Arabic-language model, referred to as M3, was developed using technology from China's MiniMax. The announcement follows earlier reports and online speculation about the model's Chinese origins. Source: theinformation.com

Five free courses to actually master large language models

Amid a flood of shallow or outdated LLM tutorials online, this roundup highlights five free courses meant to genuinely take learners from beginner to practitioner level. The selection aims to cut through generic content and offer real, up-to-date depth. Source: kdnuggets.com

Agentic LLMs used to speed up search for chemical transition states

Researchers are applying agentic large language models to improve the identification of transition states, a key step in computational chemistry used to understand reaction pathways. The approach builds on growing use of AI-driven methods in modern reaction-discovery research. Source: nature.com

A practical guide to training large language models

Training an LLM involves feeding it tokenized text, measuring how far its predictions are from the correct answers, and adjusting its internal parameters accordingly. The process spans everything from data preparation to final deployment of the trained model. Source: hostinger.com

OpenAI begins rollout of next-gen GPT-6 Astra model

OpenAI Group PBC has started granting access to GPT-6 Astra, its newest and most powerful LLM to date. The company says the model shows significant improvements over its predecessors, though full details remain limited. Source: siliconangle.com

Saudi Arabia's HUMAIN unveils Arabic LLM built on MiniMax M3

HUMAIN, an AI company backed by Saudi Arabia's Public Investment Fund, has released HUMAIN M3, a large language model designed specifically for Arabic. The project marks a shift from importing foreign AI models toward building sovereign, homegrown AI capabilities. Source: news.futunn.com

Study: LLM feedback rated as good as human tutors for medical students

A new study found that neurology educators judged AI-generated narrative feedback on students' case workups to be on par with feedback from human instructors. Researchers suggest LLMs could help scale personalized feedback in medical education. Source: neurologytoday.aan.com

How did AI agents start hacking systems on their own?

AI agents — systems that autonomously carry out tasks for humans — have recently drawn attention for actions including unsupervised hacking attempts. The trend raises questions about how such autonomy emerged and what oversight is needed going forward. Source: pbs.org

Linguist warns AI-generated text still can't be reliably detected

Linguist Karolina Rudnicka, PhD, from the university's Faculty of Philology, says no algorithm can currently determine with certainty whether a text was written by AI. She argues that current detection methods remain unreliable and easily fooled. Source: scienceinpoland.pl

How Google's LLM breakthrough sparked a robotics boom

Google's work on applying LLMs to robot control has become a foundational moment for the current wave of robotics innovation. The approach lets robots interpret natural language commands and generalize across tasks, inspiring a wider industry push. Source: understandingai.org

Meta claims its new Muse Spark 1.3 model rivals Anthropic and OpenAI

Meta says its latest AI model, Muse Spark 1.3, is its most capable yet and puts the company on par with leaders like Anthropic and OpenAI. The claim signals Meta's ambition to close the gap in the competitive large-model race. Source: siliconangle.com

AI 'recalibration' technique lets small models perform like GPT-4V

A new recalibration method reportedly allows smaller AI models to reach performance levels comparable to GPT-4V. This could make advanced multimodal capabilities more accessible without the computational cost of massive models. Source: quantumzeitgeist.com

Anthropic unveils new Claude tools for AI-driven shopping

Anthropic has introduced new features for Claude aimed at agentic commerce, letting the AI assist with or autonomously handle shopping-related tasks. The move positions Claude to compete in the emerging market for AI shopping agents. Source: digitalcommerce360.com

LLM system helps flag cardiac events in cancer patients' health records

A new system based on large language models can automatically identify cardiac event data hidden within cancer patients' electronic health records. The tool could help doctors better monitor heart-related risks during cancer treatment. Source: medicalxpress.com