Frontier Labs

model releases, research and announcements from OpenAI, Anthropic, Google, Meta, xAI, Mistral and the labs building frontier AI

Macro photograph of glowing amber light pulses travelling along the surface of a translucent circuit board, seen from directly above at a slight angle, deep bla
Frontier Labs

An Anthropic AI system trained itself to be safer, faster and cheaper than human researchers

A new paper from Anthropic shows an automated system that fixes AI misbehaviour on its own, outperforming experienced humans in under six hours at a fraction of the cost.

3 min read
A sleek array of glowing server racks in a modern data centre, cool blue and white light reflecting off polished metal surfaces, shot from a low angle looking u
Frontier Labs

OpenAI разрабатывает ИИ-агента, который никогда не перестаёт работать

Новый режим «Persistent mode» для Codex от OpenAI позволит ИИ-агенту выполнять задачи самостоятельно и даже отправлять вам сообщения без запроса. Вот что это означает для обычных пользователей.

4 min read
A futuristic AI model represented visually as interconnected nodes and pathways, symbolizing complex data processing
Frontier Labs

ИИ только что решил сложную открытую задачу в чистой математике. Математики потрясены.

Модель Astra компании OpenAI решила давнюю проблему теории групп в этом месяце. Работающий математик объясняет, почему это казалось невозможным еще несколько недель назад.

3 min read
Abstract lattice of light filaments suspended on near-black
Frontier Labs

Google DeepMind помещает свой ИИ в криптографически защищённый контейнер для тестирования

Первая в истории двойная слепая оценка призвана предотвратить секретное изучение моделями ИИ экзаменационных вопросов перед днём тестирования.

4 min read
A glass-walled research office at dusk
Frontier Labs

Perplexity launches Portable Computer, an AI agent that runs entirely on your own machine

Built with Nvidia, the new product lets an AI agent handle hours of document work without sending your files to the cloud or charging you per task.

4 min read
A sleek, small white cylindrical smart speaker sitting on a minimalist wooden desk near a window with soft natural light, a subtle camera lens visible on its bo
Frontier Labs

NVIDIA's New Vera Rubin Chip Does 30 Times More AI Work Per Watt Than Its Predecessor

A new benchmark puts NVIDIA's Vera Rubin NVL72 hardware far ahead on energy efficiency for the kind of complex, multi-step AI tasks that companies are rapidly building into their products.

4 min read
A vast, brightly lit data centre floor photographed from a low angle, rows of glowing server racks receding into the distance, cool blue and white light reflect
Frontier Labs

Google Launches Three New Gemini Audio Models That Handle Jargon, Interruptions and 85 Languages

Gemini 3.5 Transcribe, 3.5 Live and 3.5 Live Experimental are rolling out now, and the transcription model is a first for the Gemini family.

3 min read
A dense grid of glowing GPU server racks inside a dark data centre, cool blue and violet light reflecting off metallic surfaces, photorealistic editorial photog
Frontier Labs

OpenAI's Head of Data Centers Has Left the Company, Adding to a Long List of Senior Departures

Chris Malone, who oversaw OpenAI's data center strategy during a critical period of infrastructure growth, left last week. He is at least the 13th senior executive to exit the company in 2026.

4 min read
A sleek array of glowing server racks in a modern data centre, cool blue and white light reflecting off polished metal surfaces, shot from a low angle looking u
Frontier Labs

IBM's Granite 4.2: Three Sizes, One Big Leap Into AI Reasoning

IBM has released Granite 4.2, a family of reasoning-focused AI models that can think through hard problems step by step, use tools, and even browse the web, all for free under an open licence.

4 min read
A sleek, small white cylindrical smart speaker sitting on a minimalist wooden desk near a window with soft natural light, a subtle camera lens visible on its bo
Frontier Labs

GLM-5.3 Opens Up to Developers at the Same Price as Its Predecessor

Chinese startup Z.ai's new frontier model is now available via API, matching the old per-token rate while scoring higher on independent benchmarks. But a quirk in how the model writes means your bill may still rise.

4 min read
A futuristic AI model represented visually as interconnected nodes and pathways, symbolizing complex data processing
Frontier Labs

Grok Is Sending Users Word Salad Instead of Answers

A bug in xAI's Grok chatbot is filling some users' screens with meaningless strings of words. Here is what is happening, who is affected, and what to do if it hits you.

3 min read
Abstract lattice of light filaments suspended on near-black
Frontier Labs

Liquid AI's New Draft Models Make Its LFMs Up to 3.2x Faster, No Quality Trade-off

A technique called speculative decoding lets a small helper model do the heavy lifting so the main model just checks the work. The result: dramatically faster output on everything from a data-centre GPU to a MacBook.

4 min read
Cavernous data centre aisle
Frontier Labs

Liquid AI сжимает свои компактные модели в смартфоны и Raspberry Pi без заметной потери качества

Новая техника обучения под названием Quantization-Aware Distillation позволяет компактным языковым моделям Liquid AI работать на дешевом оборудовании, сохраняя при этом 97% точности полной точности.

3 min read
A futuristic AI model represented visually as interconnected nodes and pathways, symbolizing complex data processing
Frontier Labs

Новый глава AI в Google: одна задача — сократить отставание от OpenAI и Anthropic

Демис Хассабис уходит с поста руководителя DeepMind. Его преемник Корай Кавукчуоглу берет штурвал компании, которая, по мнению аналитиков, отстала в гонке за самыми мощными AI-моделями.

4 min read
A sleek array of glowing server racks in a modern data centre, cool blue and white light reflecting off polished metal surfaces, shot from a low angle looking u
Frontier Labs

Chinese AI Labs Now Dominate Open-Source Models. Here Is What the Numbers Actually Say.

A Hugging Face data deep-dive covering January to August 2026 finds Chinese labs releasing the biggest open models by far, while the tools developers actually depend on are years-old American ones nobody talks about.

4 min read
© 2026 AI2Day