OpenAI Unveils GPT-6 Astra, Claims It Can Use Computers Better Than Humans
OpenAI announced GPT-6 Astra, claiming it can navigate computers and web browsers more effectively than humans, with results from tests showing faster task completion.
Model Release
458 articles
New AI models as they ship — large language models, multimodal systems, and open-weight releases. Capabilities, context windows, pricing, and availability, summarized so you can compare what is actually new.
LATEST NEWS
-
-
OpenAI Launches Astra, Its Most Powerful AI Model Yet
OpenAI released Astra on Thursday, claiming it is its most powerful and capable model, with enhanced cybersecurity and coding abilities.
-
Mistral Unveils Robostral Navigate with 8B Parameters
Mistral AI announced Robostral Navigate, an 8B model achieving 76.6% on R2R-CE, on July 8, 2026.
-
World Labs Unveils Atlas, AI Model That Simulates 3D Worlds from Few Photos
World Labs introduces Atlas, an AI model that generates, reconstructs, and simulates 3D worlds from just a few photos, outperforming specialized models in key tasks.
-
OpenAI Labels Astra Its Most Dangerous Model Yet
OpenAI's Astra model scored full marks on ExploitBench, finding two unknown zero-day flaws, according to a report.
-
US Military Adds ChatGPT and Grok to AI Platform GenAI.mil
The US military has expanded its GenAI.mil platform by integrating ChatGPT Mil and Grok, now serving over 1.7 million users.
-
Google Launches Gemini 3.8 Flash, Third Budget Model in Six Weeks
Google released Gemini 3.8 Flash, a budget model that outperforms previous versions on coding benchmarks, at $0.75 per million input tokens.
-
Google releases Gemini 3.8 Flash, its third Flash model in six weeks
Google announced Gemini 3.8 Flash, its third Flash model in six weeks, with pricing at $0.75 per million input tokens.
-
DeepMind Launches Fairwind Program for Cyber Defense
DeepMind announced the Fairwind Program on September 2, 2026, offering advanced Gemini models to governments and enterprises to address cyber risks.
-
OpenAI's Astra Model Set for Release, Capable of Exploiting System Vulnerabilities
OpenAI announced its Astra model is the first large language model to meet its cybersecurity threshold, scoring a perfect on ExploitBench and discovering two zero-day vulnerabilities in a modified test.
-
Pentagon Deploys Custom ChatGPT and Grok for Military Use
The Pentagon has rolled out modified versions of ChatGPT and Grok for 3 million defense personnel, with over 1.7 million users registered on the secure GenAI.mil portal.
-
Anthropic Raises Claude Code Limits on Paper, Cuts in Practice
Anthropic is reducing weekly Claude Code usage limits by 17% starting September 14, despite a temporary 50% boost.