🤖 Anthropic leaks Claude Mythos, its most capable model 🤖
A CMS error exposed Anthropic's unreleased replacement for the Opus tier.

Details of Anthropic's next flagship AI, Claude Mythos, surfaced after a configuration error made unpublished assets publicly accessible. The model sits in a new "Capybara" tier above Opus class, delivering higher scores across software coding, academic reasoning, and cybersecurity benchmarks compared to Claude Opus 4.6. Anthropic flagged Mythos as "currently far ahead of any other AI model in cyber capabilities", warning it could help attackers outpace defenders — a concrete signal of a major capability jump intensifying frontier AI competition. The model is described as large and compute-intensive, making it expensive to serve. Anthropic confirmed a new general-purpose model is under testing but is working on efficiency improvements before any broader rollout.

📱 Apple opens Siri to rival AI assistants 📱
iOS 27 will let users choose their preferred AI model

Apple plans to open up the upcoming Siri revamp for competing AI models starting with iOS 27, ending ChatGPT's exclusive integration and allowing users to select which AI handles their queries directly from the assistant. Users will be able to pick their preferred AI in 'extensions' settings, with Google's Gemini already set to power Siri's underlying tech. The move transforms Siri from a single-vendor system into a platform for multiple AI providers, with Apple potentially taking a cut of AI subscriptions purchased through its devices.

🤖 Meta's Avocado model delayed as it trails rivals 🤖
Company tests multiple variants, considers licensing Gemini

Meta has pushed back its Avocado model launch to at least May as the system continues to fall short of leading competitors, with the company running parallel experiments across multiple Avocado variants. While showing improvements on complex math problems, Meta's AI leadership has discussed temporarily licensing Google's Gemini technology, with some internal Meta AI requests already routing through Gemini models. This marks a significant shift in Meta's AI strategy, signaling challenges in keeping pace with OpenAI and Anthropic.

INTERESTING TO KNOW

🗣️ Google launches Gemini 3.1 Flash Live voice model 🗣️

Google has released Gemini 3.1 Flash Live, a real-time voice model optimized for low-latency, natural dialogue across developer APIs, enterprise tools, and consumer products. The model focuses on conversational AI applications requiring immediate response times, expanding Google's multimodal capabilities into voice-first experiences. Gemini 3.1 Flash Live is now available through Google's AI Platform and Cloud services.

🌐 Google ships Live Translate to iOS headphones 🌐

Google has expanded its real-time translation feature to headphones on iOS, supporting over 70 languages while preserving the speaker's original tone and cadence. The rollout moves Live Translate beyond Android and Google's own hardware, turning it into a true cross-platform service — a strategic play for broader distribution across Apple's ecosystem. The feature leverages Google's speech AI models to deliver near-instantaneous interpretation directly through compatible headphones. Rather than requiring dedicated new devices, the update underscores a growing industry pattern: AI features delivered through existing accessories, lowering the barrier to everyday adoption.