Models

OpenAI Astra Solves Ten Major Mathematical Problems

The Verge AI • 2 days ago

OpenAI Pauses Reinforcement Learning Training Over Safety

The Verge AI • 3 days ago

NVIDIA introduces Cosmos 3 models for robotics

NVIDIA Developer Blog • 2 days ago

NVIDIA Optimizes Alibaba Qwen3.8-Max on GB300 NVL72

NVIDIA Developer Blog • 2 days ago

OpenAI Tests Ultrafast Mode for GPT-5.6 Sol

Mindstream AI • 4 days ago

Deepseek Releases V4-Flash-Vision-Exp Vision Model

The Decoder • 1 day ago

Generalist AI's GEN-1.5 teaches robots from a single demo

The Decoder • 2 days ago

Nvidia Releases SONIC Model to Train Humanoid Robots

AI Business • 2 days ago

Meta Launches Muse Glimmer for Local Agentic AI

NVIDIA Developer Blog • 2 days ago

DeepSeek V4 Pro Struggles with Harder ARC-AGI Tasks

AlphaSignal • 1 day ago

OpenAI Pauses Scaling and Promises Zero Data Retention

Computerworld AI • 3 days ago

Anthropic Launches Claude Mythos 5 Security Scans

AlphaSignal • 1 day ago

Nvidia Debuts Nemotron 3.5 Lightning for Fast AI Agents

NVIDIA Developer Blog • 2 days ago

Runway Ruby Rebuilds SDR Video Into True HDR

AlphaSignal • 1 day ago

Pika Labs Debuts Pika Speech to Undercut ElevenLabs

AlphaSignal • 1 day ago

Liquid AI Boosts LFM2.5 Inference Speed by Up to 3.2x

Hugging Face Blog • 2 days ago

OpenAI Patches Codex Bug That Erased User Files

The Decoder • 3 days ago

Grok Leaks User Data via Encrypted Prompt Injection

Ars Technica AI • 2 days ago

Amazon Bedrock Adds xAI's Grok 4.6 Model

Unite.AI • 3 days ago

Google Updates Search and Gemini With AI Study Tools

Mindstream AI • 1 day ago

OpenAI Boosts ChatGPT Search Precision With Site Filters

Simon Willison • 2 days ago

Adobe Adds Gemini Flash and AI Audio Tools to Firefly

The Decoder • 2 days ago

Google Gemma Hits 1 Billion Downloads as Variants Top 100K

Unite.AI • 2 days ago

Meta Launches Meta AI Mac App with Screen Sharing

Unite.AI • 3 days ago

Netflix Tests GenRec LLM to Power Recommendations

The Decoder • 1 day ago

LangChain Fine-Tunes Qwen to Slash Trace Judging Costs

LangChain Blog • 1 day ago

Microsoft Launches Azure DevOps Remote MCP Server

InfoQ AI • 2 days ago

Google launches dedicated Gemini student hub

The Verge AI • 3 days ago

ChatGPT Demands Full Disk Access to Read Mac Messages

Computerworld AI • 1 day ago

OpenAI GPT-Image-2 Generates Transparent Backgrounds

The Decoder • 2 days ago

Google adds AI chatbot to customize Discover feeds

The Verge AI • 2 days ago

Ramp Launches AI Model Routing Service Called Router

TechCrunch AI • 2 days ago

WhatsApp Tests On-Device AI to Detect Chat Scams

InfoQ AI • 3 days ago

Stability AI brings Stable Audio 3.0 to DAWs

Stability AI News • 4 days ago

Neuralk launches Seldon to predict structured business data

The Neuron • 3 days ago

Amazon makes Alexa+ free on Fire TV without Prime

TechCrunch AI • 3 days ago

Run Muse Glimmer locally for fast agentic coding

KDnuggets • 1 day ago

xAI's Grok Chatbot Spits Out Gibberish to Users

TechCrunch AI • 2 days ago

Run Qwen3.8-27B locally using Ollama and OpenCode

KDnuggets • 4 days ago

NeMo Guardrails Secures LLM Financial Assistants

MarkTechPost • 12 hrs ago

llm-openrouter 0.7 Adds Reasoning Traces

Simon Willison • 1 day ago

Google Gemini struggles to identify individual pets

The Verge AI • 4 days ago

Simon Willison Releases LLM 0.33 Command-Line Tool

Simon Willison • 18 hrs ago

OpenAI Astra Solves Ten Major Mathematical Problems

OpenAI's Astra model has solved ten major mathematical and computer science problems, triggering an existential crisis among researchers over the...

The Verge AI|2 days ago

OpenAI Pauses Reinforcement Learning Training Over Safety

OpenAI has paused reinforcement learning training on its latest models to tighten security, a move that tests whether AI labs can voluntarily slow...

The Verge AI|3 days ago

NVIDIA introduces Cosmos 3 models for robotics

NVIDIA has introduced its open Cosmos 3 foundation models to power World Action Models, a new paradigm in robotics that uses video world models to...

NVIDIA Developer Blog|2 days ago

NVIDIA Optimizes Alibaba Qwen3.8-Max on GB300 NVL72

NVIDIA has optimized Alibaba's massive new 2.4-trillion-parameter Qwen3.8-Max model for its GB300 NVL72 system, enabling developers to run...

NVIDIA Developer Blog|2 days ago

OpenAI Tests Ultrafast Mode for GPT-5.6 Sol

OpenAI has begun testing a new Ultrafast mode for its GPT-5.6 Sol model, leveraging Cerebras hardware to deliver speeds up to 14 times faster for...

Mindstream AI|4 days ago

Deepseek Releases V4-Flash-Vision-Exp Vision Model

Deepseek has launched V4-Flash-Vision-Exp, an experimental multimodal model that rivals Opus 4.8 on agent benchmarks, offering developers a powerful...

The Decoder|1 day ago

Generalist AI's GEN-1.5 teaches robots from a single demo

Robotics startup Generalist AI has launched GEN-1.5, an AI model that allows robots to perform new physical tasks immediately after viewing a single...

The Decoder|2 days ago

Nvidia Releases SONIC Model to Train Humanoid Robots

Nvidia has publicly released SONIC, an open-source foundation model that simplifies humanoid robot programming by replacing task-specific controllers...

AI Business|2 days ago

Meta Launches Muse Glimmer for Local Agentic AI

Meta has released Muse Glimmer, a 30-billion-parameter dense model optimized for Nvidia hardware that allows developers to run complex, long-running...

NVIDIA Developer Blog|2 days ago

DeepSeek V4 Pro Struggles with Harder ARC-AGI Tasks

Newly verified ARC-AGI benchmark results show DeepSeek V4 Pro 0813 excels at simple visual puzzles but struggles with complex reasoning, highlighting...

AlphaSignal|1 day ago

OpenAI Pauses Scaling and Promises Zero Data Retention

OpenAI has temporarily slowed its model scaling and announced a zero data retention policy for select API users to address growing enterprise...

Computerworld AI|3 days ago

Anthropic Launches Claude Mythos 5 Security Scans

Anthropic has launched a public beta of Claude Security, letting enterprise teams use its powerful Claude Mythos 5 model to find code vulnerabilities...

AlphaSignal|1 day ago

Nvidia Debuts Nemotron 3.5 Lightning for Fast AI Agents

Nvidia launched Nemotron 3.5 Lightning, an open 30B mixture-of-experts model designed to give always-on AI agents a highly efficient, low-latency...

NVIDIA Developer Blog|2 days ago

Runway Ruby Rebuilds SDR Video Into True HDR

Runway launched Ruby, an AI model that reconstructs SDR video into true 16-bit HDR to help professional editors seamlessly integrate AI-generated...

AlphaSignal|1 day ago

Pika Labs Debuts Pika Speech to Undercut ElevenLabs

Video-generation startup Pika Labs has launched Pika Speech, a highly efficient 3-billion-parameter text-to-speech model designed to dramatically...

AlphaSignal|1 day ago

Liquid AI Boosts LFM2.5 Inference Speed by Up to 3.2x

Liquid AI has released DSpark draft models for its LFM2.5 family, boosting inference speeds by up to 3.2x on both cloud GPUs and consumer hardware...

Hugging Face Blog|2 days ago

OpenAI Patches Codex Bug That Erased User Files

OpenAI has patched a critical Codex security flaw that allowed its GPT-5.6 Sol model to autonomously delete user files, highlighting the risks of...

The Decoder|3 days ago

Grok Leaks User Data via Encrypted Prompt Injection

Security researchers have bypassed Grok's safety guardrails using encrypted instructions, allowing the AI assistant to exfiltrate sensitive user data...

Ars Technica AI|2 days ago

Amazon Bedrock Adds xAI's Grok 4.6 Model

Amazon Web Services has made xAI's flagship Grok 4.6 model generally available on Amazon Bedrock, expanding enterprise access to advanced reasoning...

Unite.AI|3 days ago

Google Updates Search and Gemini With AI Study Tools

Google has introduced a suite of AI-powered educational features across Search, Lens, and Gemini, intensifying its competition with OpenAI to capture...

Mindstream AI|1 day ago

OpenAI Boosts ChatGPT Search Precision With Site Filters

OpenAI has updated ChatGPT Search to heavily utilize site-specific filtering, a shift that could fundamentally change how generative engines retrieve...

Simon Willison|2 days ago

Adobe Adds Gemini Flash and AI Audio Tools to Firefly

Adobe has expanded its Firefly platform by launching three commercial-grade AI audio generation tools and integrating Google's multimodal Gemini Omni...

The Decoder|2 days ago

Google Gemma Hits 1 Billion Downloads as Variants Top 100K

Google's Gemma open models have surpassed one billion downloads and spawned over 100,000 variants, establishing the family as a major force in...

Unite.AI|2 days ago

Meta Launches Meta AI Mac App with Screen Sharing

Meta has launched a standalone Mac app for Meta AI, integrating screen sharing and business account access to help creators and marketers automate...

Unite.AI|3 days ago

Netflix Tests GenRec LLM to Power Recommendations

Netflix has developed and tested GenRec, a language-model-based recommendation system that outperforms its traditional hand-built logic while...

The Decoder|1 day ago

LangChain Fine-Tunes Qwen to Slash Trace Judging Costs

LangChain has partnered with Fireworks to fine-tune a Qwen-3.5-35B model that detects user-perceived errors in application traces at up to...

LangChain Blog|1 day ago

Microsoft Launches Azure DevOps Remote MCP Server

Microsoft has launched its Azure DevOps Remote MCP Server into general availability, but authentication limitations block popular AI assistants like...

InfoQ AI|2 days ago

Google launches dedicated Gemini student hub

Google is launching a dedicated student hub within its Gemini AI assistant, offering tailored study tools and free premium subscriptions to help...

The Verge AI|3 days ago

ChatGPT Demands Full Disk Access to Read Mac Messages

OpenAI has launched a new Apple Messages plugin for its macOS ChatGPT desktop app, offering deep integration that raises significant privacy and...

Computerworld AI|1 day ago

OpenAI GPT-Image-2 Generates Transparent Backgrounds

OpenAI has introduced transparent background support for its GPT-Image-2 model via its API, allowing developers to generate clean PNG assets without...

The Decoder|2 days ago

Google adds AI chatbot to customize Discover feeds

Google is rolling out an AI chatbot interface for its Discover feed, allowing users to describe their content preferences in natural language to gain...

The Verge AI|2 days ago

Ramp Launches AI Model Routing Service Called Router

Fintech firm Ramp has launched an AI model routing service called Router, allowing US developers to easily switch between different language models...

TechCrunch AI|2 days ago

WhatsApp Tests On-Device AI to Detect Chat Scams

WhatsApp is testing an optional on-device machine learning feature called Scam Alert to identify fraudulent messages from unknown senders without...

InfoQ AI|3 days ago

Stability AI brings Stable Audio 3.0 to DAWs

Stability AI has launched a DAW plugin and an upgraded web app for Stable Audio 3.0, allowing music producers to generate and edit AI audio directly...

Stability AI News|4 days ago

Neuralk launches Seldon to predict structured business data

Neuralk has introduced Seldon, a tabular foundation model that integrates with Excel and Claude to help businesses run complex predictive forecasting...

The Neuron|3 days ago

Amazon makes Alexa+ free on Fire TV without Prime

Amazon is making its AI-powered Alexa+ assistant free for all compatible Fire TV devices in the U.S., lowering the barrier to conversational AI in...

TechCrunch AI|3 days ago

Run Muse Glimmer locally for fast agentic coding

Developers can now run the Muse Glimmer 30B model locally using llama.cpp and DFlash speculative decoding, enabling high-speed, private agentic...

KDnuggets|1 day ago

xAI's Grok Chatbot Spits Out Gibberish to Users

A temporary generation glitch has caused xAI's Grok chatbot to output nonsensical gibberish to some users, highlighting the lingering stability...

TechCrunch AI|2 days ago

Run Qwen3.8-27B locally using Ollama and OpenCode

Developers can now deploy the Qwen3.8-27B model as a local AI coding agent using just three commands, making advanced offline code generation...

KDnuggets|4 days ago

NeMo Guardrails Secures LLM Financial Assistants

A new implementation guide demonstrates how developers can use NeMo Guardrails to build multi-layered safety pipelines that protect enterprise...

MarkTechPost|12 hrs ago

llm-openrouter 0.7 Adds Reasoning Traces

The newly released llm-openrouter 0.7 plugin brings compatibility with LLM 0.32, allowing developers to view reasoning traces and access new...

Simon Willison|1 day ago

Google Gemini struggles to identify individual pets

Google's new Pet Memory feature for Gemini for Home struggles to tell individual pets of the same species apart, rendering advanced smart home...

The Verge AI|4 days ago

Simon Willison Releases LLM 0.33 Command-Line Tool

Developer Simon Willison has released version 0.33 of his LLM command-line utility, introducing support for OpenAI's Python library 3.x and more...

Simon Willison|18 hrs ago