🧠 Large Language Models

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

1,700 articles

Google Cloud debuts Gemini for healthcare

Google Cloud has announced Gemini for healthcare, an AI solution meant to bring together fragmented clinical, operational and administrative systems. Source: blog.google

Google unveils Gemini 4 Carbon, a model built for developer agents

Developers can start building with Gemini 4 Carbon from today. Google calls it its strongest model, with frontier performance in cybersecurity and long-horizon tasks. Source: blog.google

Nvidia's GR00T 2 learns robot movements by watching human videos

Nvidia's new AI model GR00T 2 is meant to reduce training effort by transferring human movements from video footage to robotic arms. Source: the-decoder.com

OpenAI turns ChatGPT into a business directory with AI agents

OpenAI is opening ChatGPT to businesses through a new Business Hub, where companies can claim their profile, add product data and track how users engage with them. Source: the-decoder.com

Anthropic kicks off IPO roadshow with $70–$78 price range

Anthropic has launched the roadshow for its proposed initial public offering, with shares expected to price between $70.00 and $78.00. The company has applied to list its Class A common stock on the Nasdaq. Source: anthropic.com

LLM erases a concept from a diffusion model by editing weights directly, no gradients

Researchers tested whether a general-purpose LLM, given tool access to the weights of a text-to-image diffusion model, can remove a concept without any gradient-based training. The approach relies on direct weight editing instead of retraining. Source: alphaxiv.org

PayPal lifts AI accuracy from 43% to 95% without changing its LLM

PayPal says internal question-answer accuracy jumped from 43% to 95%. The company achieved this without swapping out the underlying large language model. Source: venturebeat.com

Eight countries and leading AI labs agree on shutdown mechanisms in Paris declaration

Eight countries and top AI labs have agreed to shutdown mechanisms in a new declaration signed in Paris. The news comes amid clashes between U.S. companies and governments over AI oversight. Source: the-decoder.com

ChatGPT adds verified AI 'Business Agents'

ChatGPT will let users chat with verified 'Business Agents' that represent brands and help with tasks such as comparison shopping. Source: theverge.com

Anthropic eyes IPO valuation of up to $1.4 trillion

The expected IPO price of $70.00 to $78.00 per share would value Anthropic at as much as $1.4 trillion at the top of the range. Source: theverge.com

KAIST's CluSTER cuts LLM fine-tuning time by nearly 70%

Researchers at KAIST have built CluSTER, a method that balances how data is selected and allocated during fine-tuning. It reduces large language model fine-tuning time by up to 69.6%. Source: bioengineer.org

Google launches Multimodal Knowledge Agent benchmark

Google has introduced the MKA benchmark, designed to test how well agents reason through complex visual tasks. The accompanying report shows how Gemini 4 Carbon performs. Source: blog.google

Trump touches on AI safety and China in 60 Minutes interview

In a 60 Minutes interview, Trump said the DOJ is handling AI safety risks and claimed China has 'no guardrails'. Source: theverge.com

Nemotron-Personas-Finance-Japan: synthetic dataset for Japanese financial AI

The dataset broadens the environment in which Japanese developers and enterprises can build and evaluate AI tailored to local and industry-specific needs. Source: huggingface.co

Hugging Face explores decision models beyond text generation

Large language models excel at generating text, from answering questions and writing code to reasoning about problems. The article looks at decision models as a step beyond that. Source: huggingface.co

Satirical app pings Sam Altman daily about OpenAI's alleged RAM stockpile

A tongue-in-cheek app sends Sam Altman a daily message asking whether it can also report how much RAM OpenAI has supposedly stockpiled in warehouses to keep it off the market. The joke plays on public frustration over memory supply and OpenAI's appetite for hardware. Source: tech.yahoo.com

OpenAI introduces Businesses in ChatGPT

Companies can now manage their business information in ChatGPT, see how they appear to users, and turn interest into action. Source: openai.com

Study evaluates PII protection for AI processing Japanese business documents

AI applications handling Japanese business documents must protect personally identifiable information (PII) while keeping the documents' meaning intact. The work evaluates how well this balance can be achieved in practice. Source: huggingface.co

Law firm Gilbert + Tobin rethinks legal work with AI

The firm describes its AI journey in three phases: experimentation, personal productivity and organisational productivity. The interview explores how it all began. Source: openai.com

How well does an LLM agree with the Revised Geneva Score on synthetic patients?

A computational concordance study checked how a frontier LLM categorizes pulmonary embolism risk across synthetic patient profiles. Its outputs were compared against the Revised Geneva Score. Source: cureus.com