- The Rundown AI
- Posts
- Google’s Gemini update climbs the leaderboard
Google’s Gemini update climbs the leaderboard
PLUS: HeyGen’s Avatar IV brings new emotion to AI
Zach Mink & Rowan Cheung
May 07, 2025
Good morning, AI enthusiasts. The AI leaderboard just got another shakeup, with Google dropping a new Gemini 2.5 Pro version right ahead of its I/O event and taking coding and web development to the next level.
Google may have just taken the next step up the AI ladder — but with GPT-5, DeepSeek R2, and other powerhouse models waiting in the wings, the next generation of rivals may not be far behind.
In today’s AI rundown:
Google’s Gemini 2.5 Pro climbs leaderboards
HeyGen brings new emotion to animations
Create a personal financial assistant with Zapier Agents
Lighttricks’ new open-source video model
4 new AI tools & 4 job opportunities
LATEST DEVELOPMENTS
Image source: Google
The Rundown: Google just released an early preview of Gemini 2.5 Pro I/O Edition, an update that dramatically improves coding and web development capabilities — pushing the model to the top spot across the AI leaderboard rankings.
The details:
The update achieved the top score on the WebDev Arena leaderboard, surpassing the previous frontrunner, Claude 3.7 Sonnet, by a significant margin.
The model brings enhanced performance for frontend and UI development, code transformation, editing, and creating sophisticated agentic workflows.
2.5 Pro also features new video understanding capabilities, enabling workflows like converting video content into interactive learning applications.
In addition to coding, the model takes the No. 1 spot across all categories on the LM Arena leaderboard, beating OpenAI’s o3.
Why it matters: Google’s anticipated I/O event is still weeks away, but the tech giant couldn’t wait to flex its new powerhouse to the world. Much like December’s quiet barrage of SOTA upgrades, Google continues to ship top models without the hype. If the demos and early tests are any indication, vibe coding just leveled up in a big way.
TOGETHER WITH GAMMA
The Rundown: Gamma makes creating AI-driven presentations, social posts, and websites feel like magic — turning your prompts into professional content in seconds with no design skills (or headaches) needed.
Using Gamma's intuitive platform, you can:
Transform simple prompts into complete, visually stunning presentations in minutes
Turn boring docs into attention-grabbing content that actually gets read
Build social media carousels, microsites, and more with no design or coding skills
Collaborate live and maintain brand consistency with team-focused AI tools
Visit gamma.app today and join 50M other users quickly creating standout content.
HEYGEN
Image source: HeyGen
The Rundown: HeyGen just rolled out Avatar IV, a new AI model capable of creating lifelike and expressive animations from a single photo while capturing vocal nuances, natural gestures, and facial movements.
The details:
A new diffusion-inspired ‘audio-to-expression’ engine analyzes voices to create photorealistic facial motion, micro-expressions, and hand gestures.
The model requires just a single reference image and a voice script, and works with shots like side angles and various subjects like pets and anime characters.
Avatar IV also supports portrait, half-body, and full-body formats, allowing for more dynamic and non-traditional video generations.
HeyGen said the new model excels for videos, including influencer-style UGC, singing avatars, animated game characters, and expressive visual podcasts.
Why it matters: HeyGen continues to build on creating AI avatars that are virtually indistinguishable from reality, but new support for different camera shots and formats opens up completely new workflows that break free from the typical “talking head” avatars we’ve grown used to in AI generations.
AI TRAINING
The Rundown: In this tutorial, you will learn how to create an AI-powered system that automatically extracts information from invoices in Google Drive, categorizes expenses, and organizes everything in a Google Sheet for easy financial tracking.
Step-by-step:
Visit Zapier Agents, click the plus button, and create a New Agent
Click “Configure,” name your agent, and select “Add Behavior”
Set up Google Drive as the trigger for when a new invoice is uploaded and add three tools: Google Drive to retrieve the file, ChatGPT to extract invoice data, and Google Sheets to add the information to your spreadsheet
Test your agent and toggle it “On” to activate
Pro tip: Create a dedicated folder named “Invoices” in Google Drive to have the agent only look for new files there. Please make sure to double-check its responses as AI can hallucinate.
PRESENTED BY VANTA
The Rundown: AI is rewriting the rules for healthcare companies— and also raising the stakes for data security. Join Vanta, Modern Health, and US Med-Quip for an interactive panel on how leading teams are streamlining and scaling compliance with limited resources.
This discussion will explore:
Real-world approaches to protecting sensitive patient data while adopting AI
How to build scalable compliance frameworks with limited resources
Practical ways to turn regulatory requirements into trust-building opportunities
Future-proofing your compliance strategy as regulations evolve
LIGHTTRICKS
Image source: Lighttricks
The Rundown: Lightricks unveiled LTXV-13B, an open-source AI video generation model that creates high-quality content 30x faster than existing models while being efficient enough to run on standard consumer hardware.
The details:
The model uses “multiscale rendering,” a new approach that creates videos in layers of detail, allowing for smoother and more consistent renderings.
It’s also able to run on everyday consumer GPUs while maintaining speed and quality, removing the need for expensive, enterprise-level computing power.
New features include precise camera motion control, keyframe editing, and multi-shot sequencing tools for professional-quality results.
LTXV is open source with free licensing for companies < $10M in revenue, and backed by partnerships with Getty Images and Shutterstock for training data.
Why it matters: There is no shortage of mind-blowing AI video models on the market, each making the options from just a year ago look like a completely different era of tech. With motion, consistency, speed, and efficiency continuing to scale and a flurry of high-quality open-source options, video creation is quickly changing forever.
QUICK HITS
🦜 Parakeet - NVIDIA’s open-source ASR model for high-quality transcriptions
🎬 Higgsfield Effects - Cinematic VFX effects for video generations
🎥 Recraft Advanced Style Control - Mix ready-to-use styles with images
⚙️ Windsurf Wave 8 - New updates to the OpenAI-acquired coding platform
OpenAI is reportedly set to acquire coding platform Windsurf (previously named Codeium) for $3B, which would be the AI giant’s largest acquisition to date.
Google launched AI Max, a suite of features embedded into Search for advertisers to optimize and expand the reach of their campaigns.
Elon Musk’s attorney responded to OpenAI’s PBC restructuring, saying the move “changes nothing” and is a “transparent dodge that fails to address the core issues.”
Microsoft is reportedly a major holdout in OpenAI’s announced restructuring, wanting assurances that its $13.75B investment in the AI leader is protected in the new plans.
Smart ring maker OURA announced two new AI features that allow users to log their food, nutrition and monitor their glucose while receiving personalized guidance.
FutureHouse released Finch in closed beta, a new AI agent designed to handle data-driven biology analysis and discovery.
COMMUNITY
Join our next workshop this Friday, May 9th, at 4 PM EST with Dr. Alvaro Cintas, The Rundown’s AI professor. By the end of the workshop, you’ll you’ll be equipped to prompt with precision and creativity across any AI tool and improve AI responses.
RSVP here. Not a member? Join The Rundown University on a 14-day free trial.
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.
That's it for today!Before you go we’d love to know what you thought of today's newsletter to help us improve The Rundown experience for you. |
See you soon,
Rowan, Joey, Zach, Alvaro, and Jason—The Rundown’s editorial team
Reply