Meta launches OpenAI's nightmare

Huge news for Open Source! Meta just announced Llama 3.1 405b model, it beats Claude 3.5 Sonnet, GPT-4o! The best LLMs we have today.

I was pretty confident we’d have an open source model this year that beats the top closed source models, but it happend too fast!

Mark Zuckerberg in his blog also shared how open source will eventually beat closed source models. Also supported by HuggingFace CEO and a bunch of other AI researchers.

We’re finally catching up!

On top of this, Hugging Face has unveiled their Smollm models - compact powerhouses designed for devices with limited resources. Meanwhile, Elon Musk is making big moves in AI. He's building the "Gigafactory of Compute" in Memphis, aiming to create the world's most powerful supercomputer. Not stopping there, Musk's xAI is set to release Grok 2 next month, followed by Grok 3 in December, potentially giving ChatGPT a run for its money.It’s a party for open source! Let’s jump in details.

Meta releases the biggest and best open-source AI model yet

Meta has launched Llama 3.1 405B, claiming it to be the largest and most capable openly available foundation model. This model features advanced capabilities in general knowledge, math, tool use, and multilingual translation, rivaling top models like GPT-4. Alongside the 405B model, Meta has released upgraded versions of the 8B and 70B models, which support longer context lengths and enhanced reasoning.The 405B model was trained on over 15 trillion tokens using 16,000 GPUs and includes safety features like Llama Guard 3 and Prompt Guard. These models are available for download on llama.meta.com and Hugging Face, with a focus on fostering community development and feedback through the "Llama Stack" initiative.

Hugging Face Releases Smollm Models

Hugging Face has unveiled Smollm, a series of compact language models designed to outperform existing models like Qwen2 and Phi 1.5. These models are optimized for efficiency, allowing them to run effectively on devices with limited computational power while maintaining high accuracy in various natural language processing tasks. Initial benchmarks show that Smollm excels in performance, showcasing advancements in training techniques and model architecture. By providing these models as open-source, Hugging Face aims to democratize AI, enabling developers and researchers to leverage cutting-edge technology without significant investment. The versatility of Smollm makes it suitable for a wide range of applications, from chatbots to content generation, reinforcing Hugging Face's commitment to innovation and accessibility in the AI community.

Elon Musk building worlds powerfull Supercomputer

Elon Musk's xAI is set to construct the world's largest AI supercomputer, named the "Gigafactory of Compute," in Memphis, Tennessee. This ambitious project, which has been described as the largest capital investment by a new-to-market company in the city's history, is powered by 100,000 Nvidia H100 chips and is expected to achieve a processing speed of 2.5 exaflops, more than double that of the current record-holder, the US Department of Energy's Frontier computer. Musk anticipates that this supercomputer will be operational by December and capable of training advanced AI models, including a new version of xAI's Grok chatbot, in just a week. However, the project has raised concerns among local environmentalists due to its substantial energy requirements, potentially needing as much electricity as 100,000 homes during peak operations. Despite these challenges, the investment is expected to create numerous high-quality jobs and position Memphis as a hub for AI innovation, contributing to ongoing debates in the tech community about the future of AI capabilities and the role of supercomputers in advancing the field.

xAI to Launch Grok 2 and Grok 3

Elon Musk's xAI is set to release Grok 2 next month, followed by Grok 3 in December. These updates aim to enhance the capabilities of the Grok chatbot, which is designed to compete with other advanced AI models like OpenAI's ChatGPT. The upcoming versions are expected to feature improved performance and functionality, reflecting xAI's commitment to advancing AI technology. The releases are part of Musk's broader vision for xAI, which includes leveraging powerful computing resources to develop cutting-edge AI applications.

Hand Picked Video

In this video, we'll guide you through the process of installing Meta's powerful language model, Llama 3.1, locally on your computer.

Top AI Products from this week 

• Supermemory - You've been collecting bookmarks on the internet for all this while- it's finally time to use them. Supermemory is a hub for organizing and utilizing saved information, with a search engine, writing assistant, canvas and more.

• Reactor Chat AI - I'm Reactor, ARC's creation, victor over GPT4o in MMLU and HumanEval! I'm speedy, eco-friendly (only 0.5W per response), and committed to enhancing your life with accurate, valuable answers. Let's be friends!

• Dialog AI - Conduct real-time, interactive interviews covering technical, coding, and phone screens. Our AI ensures a consistent and fair process and then creates a customized report after each interview, giving you the insights needed to make confident hiring decisions.

• Vozo Rewrite & Redub - Forget the traditional video editing! Vozo enables users to rewrite, redub, edit voice, and lip-sync narrative videos into new stories with prompts.

• Shaped - Increase engagement, conversion, and revenue with configurable recommendations and search that adapt in real-time

• Gobi - Gobi is a human-quality wellness instructor that deeply understands your needs and generates science-backed, voice-guided wellness practices for your emotional well-being in real-time.

• HackerPulse - Bring together your GitHub, Stack Overflow, and LinkedIn profiles into a single HackerPulse dev profile. Perfect for job applications and networking, HackerPulse profile gives a complete picture of your skills and projects, all in one place.

• Arcane - Get inspired by the latest trends, news, and audience conversations delivered by our Researcher Apps. Turn insights into valuable content, craft thought leadership posts with your TOV and re-purpose long-form content in one platform.

This week in AI

• NeuralGCM Hybrid Weather and Climate Forecasting Mode - Meta AI researchers developed NeuralGCM, a hybrid general circulation model that blends differentiable atmospheric dynamics solvers with machine learning. It delivers accurate 5-day weather forecasts, matching top models, with computational savings and capturing emergent phenomena. NeuralGCM represents a major advancement in combining machine learning and physics-based modeling for weather and climate prediction.

• Luma AI's New Looping Feature - Luma AI has launched "Loops" in its Dream Machine platform, enabling users to create seamless, five-second looping videos from text or image prompts. This feature enhances AI-generated animations, making them more engaging and fluid. It's particularly beneficial for digital marketers and content creators, allowing for quicker production of captivating video content.

• Mistral’s Codestral Mamba New Code Generation Model - Mistral has released Codestral Mamba, a 7B-parameter code model with a 256k token context window, enabling fast local code generation. It outperforms similarly sized models but is still surpassed by the larger Codestral.

• Apple Introduces Advanced AI Models - Apple has unveiled new AI models that surpass those from Mistral and Hugging Face, showcasing the company's enhanced capabilities in artificial intelligence. These models aim to improve user experiences across its product ecosystem.