This is the fifth model in the ThinkingCap series. We reduced how much Google's gemma-4-26B-A4B-it thinks, without compromising the quality of its answers.

The Pentagon has given the Department of Homeland Security permission to test a military-grade heat ray and develop tactics to potentially use it in operations at the southern US border, according to a memo signed by Defense Secretary Pete Hegseth last month…

Kimi Code Documentation

Most people know Sonic 06 has cut content, but few realise just how much...

If you take a scene from a horror movie and add funny music, the killer suddenly looks hilarious. Isn’t it strange that that happens?

Introducing Kimi K3.1 Pro, our most capable 3T-class open-weight model with 1M-token context and major advances in coding, 3D modeling, and financial research.

When timestamps aren’t time-ordered and other oddities

(BI) Software is dead, apparently

"Like all tools, it has its pros and cons. It brings about radical changes, but we'll adapt." This cliché was true for previous technologies. But it isn’t necessarily true for AI. A tool extends the human body or percept...

Analysis of Kimi's Kimi K3.1 Pro and comparison to other AI models across key metrics including quality, price, performance (tokens per second & time to first token), context window & more.

White noise machines are widely used for sound masking, increasing privacy in offices and helping people sleep. The classic approach is a fan in a box, though in the past ten years people have mostly switched over to electronic ones. But a fan is really…

You learned to talk by listening. So did Dessa. A research preview of a full-duplex speech model.

Kimi K3.1 Pro scores 55.1 on the Artificial Analysis Intelligence Index, 2.5 points short of Claude Opus 5.5: open-source AI is now 31 days behind the best closed model, down from 104. It extends China’s run: a Chinese lab has had the best open model on every…

Somehow a film made 56 years ago named Colossus The Forbin Project predicted the world of today

After the Spelling Society's London congress, Anatoly Liberman makes the case for reforming English spelling slowly: drop needless letters in words like acquire, communal, and sufficient, offend no one, and let popular culture help.

According to the new terms, sharing your data for better AI recommendations also means opening it up for more personalized ads.

OpenAI explains how we strengthened containment for frontier model training and evaluation, including offline environments and high-compute red-teaming.