Here is today's AI news. Story 1. OpenAI. GPT-6 Astra launched on Sept. 3, 2026, with OpenAI calling it its most capable model yet and emphasizing new cybersecurity and computer-use abilities.. OpenAI launched GPT-6 Astra on Sept. 3, 2026, rolling it out first to a limited set of organizations. OpenAI called it its “most capable broadly deployed model” and highlighted computer-use, coding, and cybersecurity gains. The phased release matters because it puts agentic AI into enterprise workflows while adding stricter safety gates around advanced cyber tasks. Story 2. Anthropic. Claude Fable 5.1 and Claude Mythos 5.1 shipped in the first week of September, with Anthropic highlighting major gains for coding and knowledge work plus lower cache-read pricing.. Anthropic launched Claude Fable 5.1 on September 1, 2026, calling it its most capable model for coding and knowledge work. The same underlying model, Claude Mythos 5.1, is available only to vetted organizations through Project Glasswing. Both support a 1-million-token context and 128,000-token outputs. Anthropic also cut cache-read pricing 75%, to $0.25 per million tokens, potentially lowering costs for long-running software and research workflows. Story 3. Google. Google released Gemini 3.8 Flash in early September 2026, positioning it as a faster, cheaper Flash update aimed at coding and agentic tasks.. Google launched Gemini 3.8 Flash on Sept. 2, 2026, its GA “workhorse” model for long-horizon software engineering and autonomous agents. Google says it improves on 3.7 Flash while retaining Flash speed and cost: an introductory $0.75 per million input tokens and $3.75 per million output tokens through 2026. It supports text, image, video, audio and PDF inputs with a 1M-token context window. Story 4. Meta. Meta rolled out Muse Spark 1.3 on Sept. 2, 2026, promoting stronger frontier-style performance for coding and agentic workflows at aggressive pricing.. Meta launched Muse Spark 1.3 on Sept. 2, 2026, making it available in Muse Code and the Meta Model API. The update targets coding and agentic workflows, with Meta saying it delivers “~20% fewer tool calls” and “~25% fewer tokens” than 1.2. It matters because Meta is pushing frontier-model performance at aggressive pricing while expanding access to developers and, soon, Instagram/Facebook users. Story 5. Microsoft. Microsoft faced major AI-service disruption tied to an Azure data center issue in early September 2026, temporarily impacting Copilot and other Azure-backed AI products.. Reports of a Sept. 3, 2026 Azure East US failure briefly disrupted AI services including Copilot, ChatGPT, Claude and Grok. However, Microsoft’s Azure history showed no confirmed regional incident or root cause. The event underscores AI’s dependence on concentrated cloud infrastructure: one suspected ingress/connectivity fault can cascade across competing products. That's all for today's AI news briefing.