🧠 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

Add LLM translation to scikit-learn pipelines with Scikit-LLM

A tutorial shows how to build a two-stage scikit-learn pipeline for multilingual text classification. The pipeline is driven by two pre-trained language models via Scikit-LLM. Source: kdnuggets.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

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

Why telcos can't afford to ignore decision APIs

Five of tech's biggest names launched decision APIs within roughly six weeks. The open question is whether these tiny, ultrafast AI models are the next edge play for telecom operators. Source: fiercewireless.com

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

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

Why AI companies are making AI cute

The piece asks why AI firms are giving their products a friendly, cute personality. The snippet itself is mostly a list of related Emad Mostaque video titles, including claims about economic life expectancy and Claude use cases, offering little substance on the main question. Source: youtube.com

ChatGPT Enterprise lands in AWS Marketplace

Organizations can now buy ChatGPT Enterprise through AWS Marketplace. The listing lets companies fold the purchase into their existing AWS procurement and billing processes. Source: openai.com

Google adds TPU 8tx and 8i Max for the agentic era

Google has widened its eighth-generation TPU lineup with two new chips, TPU 8tx and TPU 8i Max. Together they bring the number of new TPU systems launched within a single year to four. Source: blog.google

Meta's Muse AI agent: what we know so far

Meta has launched Muse, an AI agent it says can handle tasks ranging from sending emails to shopping online. Early impressions suggest it can be surprisingly effective, even if a bit unsettling. Source: theverge.com

Prompts may carry hidden baggage in modern AI models

A post from the Repligate circle argues that prompts can transmit more information than users expect when fed to modern models. The discussion, shared via @kromem2dot0 and @repligate, follows up on an earlier promised write-up. Source: digg.com

Anthropic unveils Claude Fable 5.2 and Mythos 5.2

Anthropic announced two new models, Claude Fable 5.2 and Claude Mythos 5.2. The company says they are built to tackle especially demanding technical challenges. Source: theverge.com

Joke about Opus simulating an explanation of itself goes around

A post shared by @repligate and @Lari_island jokes about Opus simulating its user explaining Opus back to itself, until an even deeper inner voice kicks in. The post is about three weeks old. Source: digg.com

Pliny the Prompter's Hugging Face profile shows new activity on Obliteratus

The Hugging Face profile page tied to Pliny the Prompter lists recent activity, including a discussion on the obliteratus project asking to restart it. It also references the Huihui-GLM-5.2-abliterated-GGUF model from huihui-ai. Source: huggingface.co

Google Maps Platform now built into AI Studio

Developers can now use Google Maps Platform directly inside Google AI Studio. This makes it much easier to add real-world context to apps without extra setup. Source: blog.google

HOPE: a smarter way to prune experts in MoE models

A new paper introduces HOPE (Higher-Order Pruning of Experts), a second-order method that accounts for pairwise interactions between experts. The approach aims to preserve model quality even at aggressive pruning levels such as 50%. Source: huggingface.co

Kuken Industries searches 70 years of drawings in minutes

Engineers at Kuken use a tool built on the OpenAI API to locate drawings for replacement parts. A search that once took nearly an hour now takes about 4.5 minutes. Source: openai.com

BMC Urology study pits three LLM systems against clinical management cases

A blinded, case-based study published in BMC Urology compared the clinical management recommendations of three large language model systems against other comparators. The work evaluates how well such systems perform on real urological cases. Source: bioengineer.org

Control Room: a central system to track and halt model workloads

Control Room is a central system for monitoring model workloads across the research infrastructure. It allows them to be stopped quickly when safety or security concerns arise. Source: openai.com

Red Hat ships llm-d 1.0 for distributed LLM inference on Kubernetes

Edoardo Vacchi and other Red Hat engineers have released llm-d 1.0. The project routes and schedules large language model workloads across Kubernetes clusters. Source: news.lavx.hu