Foundation Models - News Articles

Compression and Prediction Are One: The Math Behind How AI Language Models Think

Aug 12, 2026

ngrok blog

A groundbreaking mathematical link between data compression and AI prediction reveals that language models are essentially compression engines, a discovery with major practical implications for making LLMs faster and more efficient through techniques like quantization.

Research Foundation Models

Nvidia Launches 30B AI Model and Open-Source Routing Tool That Slashes Inference Costs by Up to 74%

Aug 12, 2026

Venturebeat

Nvidia launches Nemotron 3.5 Lightning, a powerful 30-billion-parameter open AI model, paired with NeMo Switchyard, an open-source routing tool that dynamically assigns tasks to the most cost-efficient model available, slashing inference costs by up to 74% while maintaining frontier-level performance.

Foundation Models Open Source Coding & Developer Tools Agents

Google DeepMind Gets New SVP as Tech Giant Races to Close AI Gap with OpenAI and Anthropic

Aug 12, 2026

CNBC

Google DeepMind appoints Koray Kavukcuoglu as SVP to lead Gemini development as the tech giant races to close a growing AI performance gap with OpenAI and Anthropic, signaling a major strategic push toward faster model releases and product-focused execution.

Foundation Models Business

Oumi Launches Platform Letting Enterprises Build and Own Custom AI Models in Hours, Challenging Big AI Labs

Aug 12, 2026

The Deep View

Oumi, founded by a former Google Gemini lead, launches a platform enabling enterprises to build and own custom AI models in hours, challenging Big AI Labs as major companies like Thomson Reuters ditch costly rentals from OpenAI and Anthropic in favor of full control, customization, and data privacy.

Enterprise Foundation Models Open Source

Nvidia Launches Free Lightweight AI Model That Runs on a Single GPU, Taking Aim at OpenAI and Anthropic

Aug 12, 2026

CNBC

Nvidia launches Nemotron 3.5 Lightning, a free, open-source AI model that runs on a single GPU, directly challenging OpenAI and Anthropic while companies like CrowdStrike and Harvey are already putting it to the test.

Foundation Models Open Source Hardware Agents

NVIDIA Launches Nemotron 3.5 Lightning: 30B Parameter AI Model Runs 30% Faster Than Competitors With Only 3B Active Parameters

Aug 12, 2026

NVIDIA Technical Blog

NVIDIA launches Nemotron 3.5 Lightning, a 30B parameter open Mixture-of-Experts AI model that activates only 3B parameters at a time, delivering up to 4x faster output speeds and completing 10,000 agentic tasks 30% faster than Qwen3 35B, with fully open weights released under OpenMDW-1.1 for deployment anywhere from local hardware …

Foundation Models Agents Open Source Hardware

Google's Gemini Hits 1 Billion Users But Faces Internal Turmoil Amid Competitive Pressure

Aug 11, 2026

Ars Technica

Google's Gemini AI reaches 1 billion monthly active users, becoming the fastest-growing product in company history, but faces mounting internal pressure as key researchers depart, model releases are delayed, and coding performance continues to trail rivals OpenAI and Anthropic.

Foundation Models Business Labor & Jobs

Meta Releases Muse Glimmer: A 30-Billion-Parameter Open-Weight AI Model Built for Autonomous Agents

Aug 11, 2026

Venturebeat

Meta launches Muse Glimmer, a powerful 30-billion-parameter open-weight AI model under Apache 2.0, designed for autonomous agents, capable of running on consumer hardware with 24GB VRAM, and outperforming key rivals on several agentic benchmarks.

Open Source Agents Foundation Models

OpenAI Launches Two-Tier AI Cyber Defense Service as AI-Driven Attacks Surge

Aug 11, 2026

TechCrunch

OpenAI launches a two-tier AI cyber defense service called Daybreak, featuring a specialized GPT-5.6 Cyber model for security testing, even as critics question whether AI labs are profiting from the very cyber threats their technology helps create.

Security Foundation Models Business

Meta Launches 30-Billion Parameter AI Model Free for Local Use as Zuckerberg Warns Against AI Power Concentration

Aug 11, 2026

The Deep View

Meta launches Muse Glimmer, a powerful 30-billion parameter open-weight AI model that runs locally on consumer hardware, as CEO Mark Zuckerberg publishes a sweeping manifesto warning against AI power concentration — even as critics question whether Meta itself represents exactly that.

Open Source Foundation Models Agents