๐Ÿš€ Google Gemini Tops AI Charts

PLUS: Flux Image AI Beats Stable Diffusion

Welcome back!

Google just shook up the AI world. A new version of Gemini 1.5 Pro now leads the Chatbot Arena Leaderboard, outshining GPT-4o and Claude-3.5 Sonnet. Meanwhile, their tiny Gemma 2B model outperforms GPT-3.5 and Mixtral 8x7B. Bigger is not always better. Let's unpack this...

Todayโ€™s Summary:

  • ๐Ÿš€ Google Gemini tops AI charts

  • ๐ŸŽจ Flux Image AI beats Stable Diffusion

  • ๐ŸŒ Google adds AI to Chrome

  • ๐Ÿ› ๏ธ New Stability 3D model

  • ๐Ÿ›๏ธ EU AI Act activated

  • ๐ŸŽฅ Runway faster Turbo model

  • ๐Ÿค– Zuckerberg hints at Llama 4 scale

  • 2 new tools

TOP STORY

New Google Gemini model beats GPT-4o and Claude Sonnet

The Summary: Google Gemini 1.5 Pro has claimed the top spot in the Chatbot Arena Leaderboard, outperforming GPT-4o and Claude-3.5 Sonnet with an impressive score of 1300. This breakthrough marks the first time a Google model has led the leaderboard. Gemini 1.5 Pro shows exceptional multilingual capabilities and strong performance in technical domains. Additionally, Google released a tiny Gemma 2 model beating GPT-3.5.

Key details:

  • Gemini 1.5 Pro Experimental-0801 scores 1300 in Chatbot Arena, surpassing GPT-4o and Claude-3.5 Sonnet

  • Ranks #1 in Overall and Vision categories, excelling in multilingual tasks

  • Strong performance in Math, Instruction-Following, and Coding

  • Tiny Gemma-2-2B outperforms GPT-3.5 and Mixtral 8x7B

  • ShieldGemma for content safety filtering and Gemma Scope for model interpretability

Why it matters: Google models are reshaping the competitive landscape. Gemini 1.5 Pro dominance in various categories signals a potential shift in AI leadership. Meanwhile, the tiny Gemma 2 open-source model challenges the "bigger is better" approach.

IMAGE AI

Open source Flux image AI beats Stable Diffusion, rivals Midjourney

The Summary: Black Forest Labs, founded by Stable Diffusion creators Robin Rombach and Patrick Esser, launches Flux AI, a 12B state of the art text-to-image model. Flux aims to match Midjourney quality while offering open-source options. The model comes in three versions: a non-commercial dev version, a fast Apache-licensed version, and a closed-source pro version. Flux release marks a major step in accessible, high-quality AI image generation.

Key details:

  • Beats Stable Diffusion 3 Medium, rivals Midjourney

  • Supports resolutions up to 2.0 megapixels

  • Secured $31 million in seed funding led by Andreessen Horowitz

  • Team plans to release text-to-video models in the future

  • API cheaper than DALL-E 3

Why it matters: Flux is a new strong competitor to Stable Diffusion and Midjourney. It offers state of the art performance with open-source licensing. This release could push competitors to quickly improve their models.

GOOGLE

Google Brings AI tools to Chrome Desktop

The Summary: Google is rolling out three new AI-powered features for Chrome desktop: Google Lens integration, Tab Compare for shopping, and natural language search for browsing history. These tools aim to make web browsing more efficient and user-friendly. The update uses Google Gemini models to improve search capabilities, product comparisons, and history recall.

Key details:

  • Google Lens on desktop allows users to search and ask questions about anything they see on a webpage

  • Tab Compare generates AI-powered overviews of products across multiple tabs for easier shopping decisions

  • Natural language search for browsing history helps users find previously visited sites using conversational queries

  • These features use Google Gemini models and will be available in the coming weeks, starting in the US

Why it matters: These AI integrations will make web browsers more intuitive and helpful. By bringing mobile-first features like Google Lens, Google is setting new standards for browser functionality. This move could push competitors to develop similar features, ultimately benefiting users.

QUICK NEWS

Quick news

โ

โ€œThe amount of computing needed to train Llama 4 will likely be almost 10 times more than what we used to train Llama 3, and future models will continue to grow beyond thatโ€

TOOLS

๐Ÿฅ‡ New tools

  • Simply Draw - AI feedback meets art - anyone can learn to draw

  • Midjourney 6.1 - New model improves image quality, coherence, text

Thatโ€™s all for today!

If you liked the newsletter, share it with your friends and colleagues by sending them this link: https://thesummary.ai/