Devin 2.0

Reasoning models don't always say what they think

Our response to the UK's copyright consultation

PaperBench

Introducing Claude for Education

Taking a Responsible Path to AGI

Runway Gen-4

Qwen2.5 Omni: See, Hear, Talk, Write, Do It All

4o Image Generation

Gemini 2.5

Leadership Updates

Early methods for studying affective use and emotional well-being on ChatGPT

Next-generation audio models in the API

Claude can now search the web

Pixel 9a: The latest A-series phone with Google AI smarts at an unbeatable value

Gemini Canvas

BotQ: A High-Volume Manufacturing Facility for Humanoid Robots

The court rejects Elon's latest attempt to slow OpenAI down

Experiment with Gemini 2.0 Flash native image generation

Gemini Robotics

Gemma 3: The most capable model you can run on a single GPU or TPU

Gemma 3 Technical Report [pdf]

New tools for building agents

Detecting misbehavior in frontier reasoning models

Mistral OCR

Expanding AI Overviews and Introducing AI Mode

Apple introduces the new MacBook Air with the M4 chip and a sky blue color

DeepSeek-V3/R1 Inference System Overview

GPT-4.5

Amazon announces Ocelot quantum chip

Microsoft announces Phi-4-multimodal and Phi-4-mini

Amazon announces AI-powered Alexa Plus

Wan 2.1: Open and Advanced Large-Scale Video Generative Models

Claude's Extended Thinking

Apple will spend more than $500B in the U.S. over the next four years

Apple Intelligence comes to Apple Vision Pro in April

Word Deconstructor – AI-powered tool for visualizing word origins and etymology

Microsoft's Majorana 1 chip carves new path for quantum computing

Jina Deepsearch

The Anthropic Economic Index

The Intelligence Age

GitHub Copilot: The Agent Awakens

The all new le Chat: Your AI assistant for life and work

Gemini 2.0: Flash, Flash-Lite and Pro

Gemini 2.0 is now available to everyone

Updating the Frontier Safety Framework

Constitutional Classifiers: Defending against universal jailbreaks

An analysis of DeepSeek's R1-Zero and R1

Qwen2.5-1M: Deploy your own Qwen with context length up to 1M tokens

Operator research preview

DeepSeek-R1

Adebayo Ogunlesi Joins OpenAI's Board of Directors

OpenAI's Economic Blueprint

Stimulation Clicker

Why OpenAI's Structure Must Evolve to Advance Our Mission

Deliberative alignment: reasoning enables safer language models

OpenAI o1 and new tools for developers

Facts Grounding: A new benchmark for evaluating the factuality of LLMs

Imagen 3: Improved AI Image Generation

Video and image generation with Veo 2 and Imagen 3

Bringing Grok to Everyone

NotebookLM gets a new look, audio interactivity and a premium version

Gemini 2.0: our new AI model for the agentic era

Apple Intelligence now features Image Playground, Genmoji, and more

Sora System Card

ChatGPT Pro

OpenAI o1 system card

PaliGemma 2: Powerful Vision-Language Models, Simple Fine-Tuning

Genie 2: A large-scale foundation world model

Veo and Imagen 3: Announcing new video and image generation models on Vertex AI

Shazam hits 100B song recognitions

Claude for Desktop

Pushing the frontiers of audio generation

New generative AI tools open the doors of music creation

An AI Companion for Everyone

AlphaChip transformed computer chip design

Two new Gemini models, reduced 1.5 Pro pricing, increased rate limits, and more

Show HN: I made Convelyze – Visualize your ChatGPT usage with detailed analytics

Pixtral 12B

Microsoft 365 Copilot Wave 2: Pages, Python in Excel, and Agents

GameGen-O: Open-World Video Game Generation

OpenAI O1 System Card

DataGemma: Using real-world data to address AI hallucinations

AlphaProteo generates novel proteins for biology and health research

How we redesigned the Pixel 9 Pro Fold and made it thinner

1X Unveils Neo Beta, a Humanoid Robot for the Home

PearAI – Open-Source AI‑Powered Code Editor

Qwen2-VL: To See the World More Clearly

Update on Llama adoption

New in Gemini: Custom Gems and improved image generation with Imagen 3

Cerebras Inference: AI at Instant Speed

FermiNet: Quantum physics and chemistry from first principles