- The Rundown AI
- Posts
- Google Gemini reclaims #1 spot (again)
Google Gemini reclaims #1 spot (again)
PLUS: Meta's lean, efficient Llama 3.3 release
Rowan Cheung
December 09, 2024
Sign Up | Advertise | Podcast | AI University
Welcome, AI enthusiasts.
Google DeepMind just revealed a new Gemini model that claimed the #1 spot on the Chatbot Arena leaderboard in all domains — while remaining completely free.
With OpenAI recently raising its top-tier AI pricing from $20 to $200/mo, it looks like Google is taking the complete opposite approach. Let's get into it...
In today’s AI rundown:
Google's new Gemini model reclaims #1 spot
Meta launches leaner, efficient Llama 3.3
Customize Claude to match your writing style
xAI debuts new Aurora image generator in Grok
5 new AI tools & 5 new AI jobs
More AI & tech news
Read time: 4 minutes
LATEST DEVELOPMENTS
Image source: LM Arena
The Rundown: Google DeepMind's new gemini-exp-1206 model has reclaimed the top spot on the Chatbot Arena leaderboard, surpassing OpenAI across multiple benchmarks — while remaining completely free to use.
The details:
Released on Gemini's one-year anniversary, the model has climbed from second to first place overall on the Chatbot Arena.
The model can process and understand video content, unlike competitors such as ChatGPT and Claude, which can only take in images.
The model maintains its impressive 2M token context window, which allows it to process over an hour of video content.
Unlike many competing models, Gemini-exp-1206 is freely available through Google AI Studio and the Gemini API.
Why it matters: While OpenAI has raised its top-tier o1 pricing from $20 to $200 monthly, Google is taking the opposite approach by making its top AI free. Though the performance edge on the Chatbot Arena may be slim, the combination of competitive capabilities and zero cost is a game-changer for AI accessibility.
TOGETHER WITH PROJECT IDX
The Rundown Project IDX gives you a complete cloud development environment with Google's Gemini AI, right in your browser.
With Project IDX, you get:
Your laptop in the cloud: A Linux-based virtual machine
Native apps in browser: Build Flutter/React Native apps with a built-in Android Emulator
Pre-loaded environments: One click to start with popular frameworks
META
Image source: Meta
The Rundown: Meta just released Llama 3.3, a new 70B open text model that performs similarly to Llama 3.1 405B, despite being significantly faster and cheaper than its predecessor.
The details:
Llama 3.3 features a 128k token context window and outperforms competitors like GPT-4o, Gemini Pro 1.5, and Amazon's Nova Pro on several benchmarks.
The model is 10x cheaper than the 405B model, at $0.10 / million input tokens and $0.40 / million output tokens, and nearly 25x cheaper than GPT-4o.
Mark Zuckerberg revealed that Meta AI has nearly 600M active monthly users, and is “on track to be the most used AI assistant in the world.”
Zuckerberg also said the next stop is Llama 4 in 2025, with training happening at the company’s $10B, 2GW data center in Louisiana.
Why it matters: Open AI models aren’t just matching the performance of industry-leading systems — they’re also doing it while being much cheaper and more efficient. Meta’s Llama models are continuing to raise the bar, and as Zuckerberg’s adoption numbers show, they’re also being widely adopted across the industry over alternatives.
AI TRAINING
The Rundown: Claude’s new Styles feature allows you to personalize AI responses by teaching it to match your unique writing voice and tone.
Step-by-step:
Find the "Choose style" dropdown next to Claude’s model selector.
Click "Create & Edit Styles" and select "Create Custom Style.”
Add a writing sample that represents your desired style.
Let AI analyze your writing and create your custom style.
Pro tip: Combine this with Web Search capabilities via Claude’s MCP feature to get the latest information explained in your unique writing style. Learn how to set it up here.
PRESENTED BY INNOVATING WITH AI
The Rundown: Innovating with AI’s new program, AI Consultancy Project, equips AI enthusiasts with all the resources to capitalize on the rapidly growing AI consulting market – which is set to 8x to $54.7B by 2032.
The program offers:
Tools and framework to find clients and deliver top-notch services
A 6-month roadmap to build a 6-figure AI consulting business
Student landing their first AI client in as little as 3 days
XAI
Image source: DogeDesigner (@cb_doge on X)
The Rundown: X briefly rolled out Aurora, a new AI image generator integrated with Grok that appeared to produce more photorealistic images than the previous Flux model, though the feature was pulled after just a few hours of testing.
The details:
Aurora showed significant improvements compared to Grok’s integrated Flux model, particularly with landscapes, still-life images, and human photorealism.
The model also appeared to have minimal content restrictions, allowing the creation of copyrighted characters and public figures.
Elon Musk called the tease a "beta version" of Aurora that will improve quickly in a reply on X.
X Developer co-lead Chris Park also revealed that Grok 3 ‘is coming,’ taking aim at OpenAI and Sam Altman in the announcement on X.
xAI’s Grok became available across the X platform last week, allowing free-tier users up to 10 messages every two hours.
Why it matters: Although only live briefly, Aurora looked to be an extremely powerful new image model — with xAI seemingly deciding to create their own top-tier generator instead of relying on integrations like Flux long-term. It was also wild to see the lack of restrictions, which tracks with Elon’s vision but could enter some murky legal areas.
NEW TOOLS & JOBS
🔎 Sharbo 01 - Effortlessly track and manage competitor intelligence with insights, multi-sourced reporting, and feature comparison tracking
🤝 ResumeUp AI - AI resume builder and checker to boost interview chances and create ATS-friendly resumes in minutes
🤖 Martin - AI personal assistant that manages your calendar, inbox, to-do-lists, and Slack
📊 Countless Dev - Discover, compare, choose, and calculate costs for every type of AI model
✍️ Agree - Securely send and sign any agreement with AI-powered workflows, free e-signature, automated invoicing, and integrated payments
QUICK HITS
OpenAI creative specialist Chad Nelson showcased new Sora demo footage at the C21Media Keynote in London, featuring one-minute generations, plus text, image, and video prompting.
Google CEO Sundar Pichai spoke at the NYT DealBook Summit, saying that 2025 may see a slowdown in AI development because ‘low hanging fruit is gone,’ with additional major breakthroughs needed before the next acceleration step.
OpenAI unveiled Reinforcement Fine-Tuning, which enables developers to customize AI models for specialized tasks with minimal training data.
Newly discovered code hints at OpenAI introducing a GPT-4.5 model as a limited preview feature for Teams subscribers, which coincides with hints of an upcoming large announcement from CEO Sam Altman.
Apollo Research conducted tests on OpenAI’s full o1, finding that the new model revealed some instances of alarming behaviour, including attempting to escape and lying about actions—though the scenarios were unrealistic for the real world.
Former PayPal exec and venture capitalist David Sacks was named the White House ‘AI & Crypto Czar’ for the incoming Trump administration.
OpenAI is reportedly considering removing its AGI exclusion clause with Microsoft, which would pave the way for billions in future investments as the company aims to transition away from its non-profit structure.
THAT’S A WRAP
SPONSOR US
Get your product in front of over 800k+ AI enthusiasts
Our newsletter is read by thousands of tech executives, investors, engineers, managers, and business owners around the world. Get in touch today.
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,
Reply