SPEAKER_00 0:00

Welcome back to Sumo Growth AI Insights. I'm Darius, and I'm walking you through this week's AI news the way an operator actually uses it, not as a pile of headlines. Lead Story Google launches Gemini 4 Argon, its new frontier model for deep work. On Tuesday, Google made Gemini 4 Argonne globally available, calling it its latest frontier model and pitching it at long, complex work in coding, finance, legal, cybersecurity, and science. Google says Argon ranks number one on the Vals Index, which scores models on real tasks in sectors like finance, coding, legal, and tax. It supports an industry-leading 1 million token unified input and output limit through the Gemini API, with a new continuation protocol for very long jobs. Introductory pricing is $2 per million input tokens and $10 per million output tokens, with cashed input 95% off, rising to $4 and $20 after November 30th. Developers can use it now on the paid Gemini API tier in AI Studio, enterprises through Gemini for Business and Google Anti-Gravity, and consumers on Google AI Ultra and Pro Plans in the Gemini app and AI mode in search. For agencies, the intro pricing window is the moment to test Argonne on your heaviest client work, like long proposal drafts, contract reviews, and multi-step research before the price doubles. Plus Story Meta launches Muse Wonder, its first true frontier models for AI agents. On Monday, Meta announced Muse Wonder and Muse Wonderful, its most capable AI models so far, built for personal and professional agents that write code, analyze documents, and do financial research. Independent scores from artificial analysis put Muse Wonderful at 56 on its intelligence index, tied with Anthropic's Claude Sonnet 5.5, while MuseWonder scores 52, close to OpenAI's GPT-6 Astra at 53. MuseWonder is available through Meta's Model API and its Muse Code programming assistant at $4 per million input tokens and $20 per million output tokens. MuseWonderful is limited to Muse Code subscribers on the $50 and $200 monthly plans. In a big shift from its open llama days, neither model can be downloaded or self-hosted. For agencies, Meta is now one more paid frontier option to benchmark against your current model, so test it on a real client workflow before you switch anything. Three more headlines. First, NVIDIA commits $250 million in compute to UK AI agent safety research. On Wednesday, NVIDIA said it will donate compute worth $50 million a year for five years to support agentic safety and security research across the UK, including work with Imperial College London, delivered through UK-based NVIDIA Cloud Partners. Second, Microsoft unveiled Maya 300, its own AI chip, to lean less on NVIDIA. Microsoft said Maya 300 will become the primary AI inference system for its products, delivering 40% more AI output for the same power as leading GPU systems, with large-scale deployments starting in 2027. Satya Nadella called it our biggest bet yet on our own silicon. Third, OpenAI's CFO says a more powerful model is being held back for safety checks. Speaking in Dublin, Sarah Fryer said OpenAI has an unreleased internal model that proved the Navier-Stokes equations and completed 350 more math proofs last week, and that it will stay inside until it passes more safety and alignment work. Tools of the day. Kai at higherKI.ai lets you hand your calendar, email, and tasks to a proactive AI executive assistant that handles meetings and inbox work for you. MentionBird at MentionBird.ai helps you find out why AI assistants recommend your competitors and tracks how often your brand shows up in AI search answers. IrisGo at irisgo.ai lets you show it a workflow once and then runs it across your apps on its own, built for solopreneurs who want automation without code. Business AI tip of the day run a monthly AI search audit for every client. Gemini 4 Argon now powers AI mode in search for paying users, and Meta is pushing News Wonder into its own assistant. More of your clients' buyers are asking an AI, who is the best roofer near me, instead of scrolling Google results. And tools like MentionBird exist because most businesses have no idea what those answers say about them. Here's a simple monthly audit you can sell as part of every retainer. List 10 questions a real buyer would ask about the client's service and city. Ask each question in Gemini, ChatGPT, and Meta's Assistant and record which businesses get recommended. Note what the AI says about the client, services, reviews, pricing, and anything wrong or missing. Fix the gaps at the source with clear service pages, FAQ answers, fresh reviews, and consistent business details everywhere. Log the results in a custom field or note in Sumo AI CRM or Go High Level so you can show progress month over month. And trigger a review request workflow after every closed job, since recent reviews are one of the easiest signals to improve. Being recommended by AI is the new page one. The agencies that measure it first will own the conversation with their clients. Pro tip Turn the first audit into a free lead magnet. Run it for a prospect, send them a one page screenshot of competitors being recommended instead of them, and book the call from there. That's your Sumo Growth AI Insights briefing. Go see what AI says about your clients this week, and I'll catch you on the next issue.