Google introduced Gemini 4 Carbon, a new neural network that beat the flagships of OpenAI and Anthropic in the overall ranking...
🚀 Google introduced Gemini 4 Carbon, a new neural network that surpassed the flagships of OpenAI and Anthropic in the overall intelligence rating. The model received serious improvements in programming, cybersecurity and autonomous operation. According to Google, AI agents based on it are capable of days...
- 01.🚀 Google introduced Gemini 4 Carbon, a new neural network that surpassed the flagships of OpenAI and Anthropic in the overall intelligence rating
- 02.The model received serious improvements in programming, cybersecurity and autonomous operation
- 03.According to Google, AI agents based on it are capable of days...
🚀 Google представила Gemini 4 Carbon — новую нейросеть, которая обошла флагманы OpenAI и Anthropic в общем рейтинге интеллекта Модель получила серьёзные улучшения в программировании, кибербезопасности и автономной работе. По словам Google, ИИ-агенты на её базе способны днями решать сложные задачи без постоянной помощи человека. Самые интересные результаты: • $16,5 тысячи — результат в симуляции управления торговым автоматом. • 88,2% — успешность поиска уязвимостей в тесте SEC-bench Pro. • 100% отражённых атак — в конкретном тесте на защиту от промпт-инъекций. • Вдвое меньше неподтверждённых утверждений по сравнению с Gemini 4 Argon во внутренних испытаниях. Пока модель доступна только ограниченной группе специалистов по кибербезопасности. Публичный запуск обещают позже. Источник
Related Intelligence
All stories →Science Behind the Open Cyber Frontier with Mistral Large 4
The Science Behind the Open Cyber Frontier with Mistral Large 4 Ben Burtenshaw (Hugging Face) talks with Sagar Vaze and Pierre Stock from Mistral about Mistral Large 4, an open-weight model that is great for frontier cybersecurity work, amongst other tasks. They cover how the…
The ‘Pain in the Ass’ Math That Could Save Vibe Coding From Itself
A new startup called &Truth is joining the race to automate formal verification, a painstaking process that uses math to prove software works as intended.
Why does your AI website look like everyone else’s and what to connect to make it look different
Why does your AI site look like everyone else’s and what to connect to make it look different ⠀ In the video, we put together such a site through Claude Code, and under the hood most of these sites have the same set. Let’s figure out what it consists of, where to get models and ideas, and what to connect so that AI doesn’t...