Anthropic Unleashes Fable 5, Intensifying AI Battle as Apple AI Details Emerge and Ecosystem Shifts Unfold

Anthropic has officially launched Fable 5, its new flagship large language model (LLM), which, alongside a specialized, less-guarded version named Mizos 5 for cyberdefenders and infrastructure providers, sets new benchmarks across coding, agentic reasoning, and scientific research. Priced at double that of Opus ($10/million input tokens, $50/million output tokens), Fable 5 is immediately available via API, with limited complimentary access for Anthropic Pro, Max, and Team subscribers until June 22nd before transitioning to a credit-based system. This release intensifies the LLM competition, with rumors circulating about an imminent counter-release from OpenAI, potentially GPT 4.6 or 5.6. Amidst these advancements, Google DeepMind CEO Demis Hassabis emphasized that AI should augment human productivity for software developers, not replace them, a stance gaining traction as companies assess AI’s impact on their workforce. Google also announced a price drop and storage increase for Google AI Plus, alongside significant feature enhancements to its Notebook LM AI-powered research tool. The rapid progression of AI technology has, in parallel, prompted Anthropic to advocate for a temporary pause in frontier AI development, calling for a global consensus among leading AI labs to allow societal structures and research to catch up.

Apple’s recent event unveiled new AI features, dubbed Apple Intelligence, including enhanced “Clean Up” functionality in iOS 27 and a revitalized Siri. These features will be free but with usage limits, requiring an iCloud Plus subscription for expanded access. The underlying architecture sparked debate, with Apple clarifying that its “Apple Foundation Models” leverage Google Gemini technology and are powered by Google Cloud and Nvidia GPUs, supporting both local and cloud inference. Further details from the event hinted at an upcoming iPhone Fold, evidenced by new MacOS mirroring capabilities that allow resizable iPhone windows. Beyond LLM competition, broader tech discussions included a critical exposé on the Spanish railway company Renfe’s website, which allegedly performs overnight “maintenance” rendering it unusable, along with a hardcoded 3-second timeout for all page loads. Meta also addressed significant AI-driven security vulnerabilities in Instagram’s account recovery system and faced a state-level lawsuit from Florida regarding OpenAI’s “deceptive practices.” Both OpenAI and Anthropic have filed confidential S1 documents, signaling impending public stock offerings, while SpaceX, also planning an IPO, is currently valued conservatively by Morningstar despite a new $1 billion monthly compute agreement with Google.