Good morning, AI enthusiasts. While everyone is watching OpenAI’s next move, Google DeepMind just put out a reasoning powerhouse of its own.

With a massive 1M context window and breakthrough math, science, and reasoning capabilities, the bar for model releases continues to trend in one direction — up and to the right.

• Google DeepMind debuts Gemini 2.0 Flash Thinking

• Musk, Altman clash over $500B Stargate project

• Improve your content strategy with real-time AI analysis

• ByteDance releases new model, reasoning agent

• 4 new AI tools & 4 job opportunities

The Rundown: Google DeepMind just unveiled Gemini 2.0 Flash Thinking, a new free experimental AI model that sets new highs in mathematic, scientific reasoning, and multimodal benchmarks and has also moved into the No.1 spot on LM Arena’s leaderboard.

• The model achieved a 73.3% on AIME (math) and 74.2% on GPQA Diamond (science) benchmarks, showing dramatic improvements over previous versions.

• A 1M token context window allows for 5x more text processing than OpenAI’s current models, enabling the analysis of multiple research papers simultaneously.

• The system also includes built-in code execution and explicitly shows its reasoning process — with more reliable outputs and fewer contradictions.

• The model is free during beta testing with usage limits, compared to OpenAI’s $200/m subscription for access to its top reasoning model.

Why it matters: Google continues to cook — with the new Flash Thinking model beating out its own previous experimental release for the top spot on the LLM leaderboard. Plus, with reasoning capabilities and a massive 1M token context window, users are about to experience a powerhouse of intelligence and capabilities for free.

The Rundown: Perspective AI helps you have real conversations with customers at scale, uncovering the deep customer truths that help fuel business growth.

Perspective AI enables businesses to:

• Understand what drives decisions, what shapes preferences, and what your customers truly aspire to achieve

• Have natural conversations that reveal the “why” behind what customers do

• Turn deep customer understanding into actionable strategies that move the needle

The Rundown: Elon Musk just publicly challenged the funding for OpenAI’s recently announced $500B Stargate Project in a reply on X, sparking a heated exchange with CEO Sam Altman on the social media platform.

• OpenAI announced Stargate on Tuesday, with plans to build extensive AI infrastructure across the U.S. with funding from Oracle, Softbank, and others.

• Musk claimed on X that SoftBank has “well under $10B secured,” questioning the project’s financial foundation despite close ties to the Trump administration.

• Altman refuted the claims and invited Musk to the project’s first build site, also saying that Musk is “the most inspiring entrepreneur of our time.”

• Microsoft CEO Satya Nadella publicly reinforced the project’s backing in an interview on CNBC, confirming the company’s $80B commitment.

Why it matters: Following the White House Stargate reveal, many wondered what Musk’s reaction would be given his ascension into Trump’s orbit and OpenAI vendetta. We quickly have an answer — and the fight now moves into the political realm, with Elon’s role as both competitor and government advisor complicating issues even more.

The Rundown: Gemini’s Real-Time Streaming feature allows you to instantly analyze and improve your content by sharing your screen and getting immediate feedback and actionable suggestions.

1. Visit Google’s AI Studio website and select the “Share screen” feature.

2. Share your screen showing your content (YouTube channel or Instagram profile) and use prompts to analyze content strategy, design, and performance.

3. Ask for advanced optimization suggestions for immediate implementation.

Pro tip: Run real-time analysis during content planning sessions to iterate quickly. We have also written detailed prompts that you can copy and paste from here.

The Rundown: IMAGINE AI LIVE ’25 is a transformative three-day conference at the Fontainebleau Las Vegas where business leaders can connect with AI innovators and learn practical strategies for enterprise AI implementation.

Attend on May 28-30 and experience:

• Enterprise AI strategies from world-class practitioners and thought leaders

• Hands-on workshops covering practical applications of AI in business operations

• Networking opportunities with executives actively transforming their organizations with AI

The Rundown: ByteDance just released Doubao 1.5 Pro, a resource-efficient multimodal AI model that achieves high performance while using significantly fewer resources than competitors, and a new open-source reasoning agent model.

• The new model beats out rivals like GPT-4o, Claude 3.5 Sonnet, and Deepseek V3 on a series of knowledge, coding, reasoning, and Chinese benchmarks.

• Doubao’s pricing is also significantly below o1 and the new DeepSeek-R1, while claiming to use far fewer compute resources than other models.

• The company also open-sourced veRL, the reinforcement learning library used to achieve o1-level reasoning capabilities.

• ByteDance also released UI-TARS, an open-source GUI AI agent model that can reason and perform computer interactions based on screenshots for input.

Why it matters: ByteDance may have had a drama-filled few months with the TikTok ban, but it hasn’t slowed down the Chinese giant’s AI research. With DeepSeek and ByteDance dropping OpenAI-level competitors, it seems like China is having no issues keeping up in the AI race despite chip restrictions aimed at slowing them down.

• 🔎 Sonar – Perplexity’s new API for incorporating grounded, real-time search to third-party applications

• 🧊 Hunyuan 3D 2.0 – Tencent’s new open-source AI system for creating and animating high-quality 3D assets

• ❤️ Lovable + Builder.io – Turn Figma designs into full-stack apps with AI

• 📝 MeetMinutes – Transform your pre-and-post meeting workflows with AI note-taking

• 🤝 The Rundown – Partnerships Manager

• 🧳 Rad AI – Senior Executive Assistant

• 🚘 Waymo – Computer Vision/Deep Learning Engineer

• 📝 Databricks – Product Management Intern (Summer 2025)

OpenAI CPO Kevin Weil said during a presentation at Davos that the company is already training the successor to its upcoming o3 reasoning model, projecting ‘another big jump in capabilities’ despite the continuing compression of release cycles.

Microsoft modified its exclusive cloud agreement with OpenAI, retaining first-refusal rights for computing capacity while allowing the AI company to pursue additional infrastructure partnerships.

Samsung unveiled its new Galaxy S25 series phones at its ‘Unpacked’ event, featuring new AI upgrades like enhanced Gemini integrations, multimodal agent capabilities, and context-aware natural language interactions.

Google is reportedly investing an additional $1B+ into Anthropic, bringing its position to over $3B — with a separate funding deal also in the works that is set to boost the company’s valuation to around $60B.

Salesforce CEO Marc Benioff said in an interview with CNBC that the rift between Microsoft and OpenAI ‘cracked’ after Mustafa Suleyman was hired as CEO of Microsoft AI, and tensions rose between him and Sam Altman.

Ramp introduced Ramp Treasury, a new platform offering AI-powered cash management tools for businesses, including automated balance alerts, liquidity forecasting, and self-driven finance options for passive investing.

Join our next workshop this Friday at 4 PM EST to learn about how to use DeepSeek-R1 as a powerful, cost-effective alternative for your AI projects with Dr. Alvaro Cintas, The Rundown’s AI professor.

We’ll always keep this newsletter 100% free. To support our work, consider sharing The Rundown with your friends, and we’ll send you more free goodies.

Rowan, Joey, Zach, Alvaro, and Jason—The Rundown’s editorial team