- Dark Mode Digest
- Posts
- How Meta’s Llama 3.1 stacks up against ChatGPT and other models
How Meta’s Llama 3.1 stacks up against ChatGPT and other models
Zuck adds more chains to his neck in this demo of Meta AI’s latest features and other top news…
Good morning,
Today, we’ll look at the details of Llama 3.1, Meta’s latest advanced open-source LLM. Plus the top tech news you need to know from this week.
Let’s go 👇
This week’s insight
Context
This week, Meta released its latest and most advanced open-source LLM, Llama 3.1 which powers Meta AI across all its platforms and services.
Zuck published a Reel demoing some of Meta AI’s latest features, powered by Llama 3.1. He believes this will be more widely used than ChatGPT by the end of this year.
Key features of Llama 3.1
Parameter size - Available in three configurations 8B, 70B, and 405B parameters - offering flexibility for different needs.
Training data - Pre-trained on around 15 trillion tokens from publicly available sources and fine-tuned on over 10 million human-annotated examples to enhance its performance across various tasks.
Multimodal capabilities - Meta has announced plans to release multimodal features, enabling the model to process and generate not just text, but also images, videos, and other data formats.
Benchmark performance - Llama 3.1 outperforms competitors like GPT-4 in certain benchmarks, particularly in code generation tasks.
Enhanced open-source model - Designed to be accessible for researchers, developers, and organizations, promoting transparency and collaboration in AI development, unlike GPT 4 and Claude 3.5 which are proprietary
How it compares to ChatGPT and Claude’s benchmark performance
Llama 3.1 is crafted to be a formidable competitor to models like GPT-4 and Claude 3.5, all while being open-source.
Llama 3.1 compared to Nemotron 4, GPT 4, and Claude 3.5
Llama 3.1 compared to Gemma 2, Mistral, Mixtral, and GPT 3.5
Bottomline
Llama 3.1 is a big step forward in AI, giving a solid alternative to models like ChatGPT and Claude. It shines in coding and technical tasks, though there’s still work to be done in creative writing and supporting multiple languages. Being open source, Llama 3.1 makes powerful AI more accessible and could really shake up the AI world.
For more in-depth details, check out Meta’s blog.
This week’s videos
Top news
OpenAI launches SearchGPT, its AI-powered search engine, taking direct aim at Google.
Tesla’s Profit Fell 45% in the Second Quarter on Weak E.V. Sales due to slow EV sales.
Meta Unveils Largest Open-Source AI Model In History the massive Llama 3.1 405B model challenges GPT-4 and Claude 3.5 while remaining open source and freely accessible.
Other news
Spotify's price hikes didn't stop it from scoring more subscribers — and record profits where their stock was up 14% after opening bell on July 23.
Microsoft aims to boost AI workload capacity with Lumen partnership to meet growing demands at its data center.
Netflix hires former Epic Games exec as new President of Games as it’s rapidly expanding into gaming endeavors.
Amazon racing to develop AI chips cheaper and faster than Nvidia’s, executives say.
CrowdStrike blames test software for taking down 8.5 million Windows machines.
Google Abandons Plan to Phase Out Third-Party Cookies in Chrome "Instead of deprecating third-party cookies, we would introduce a new experience in Chrome that lets people make an informed choice that applies across their web browsing, and they'd be able to adjust that choice at any time."
Musk says Tesla to use humanoid robots next year to perform “unsafe repetitive or boring tasks.”
Delta canceled hundreds of flights Monday as it recovers from CrowdStrike failures.
Apple's China smartphone shipments drop 6.7% as Huawei surges, data shows.
Nvidia to launch slower version of its gaming chip in China to comply with U.S. export controls.
Google Parent’s Ad Sale Growth Slows as AI investments soar.
Introducing Llama 3.1 Meta’s most capable model to date.
Deal flow
Events startup Posh raised a $22M Series A to build the TikTok for small events. It’s now raised a total of $31M.
Vanta, a platform to automate security and compliance, raised a $150M Series C at a $2.45B valuation led by Sequoia. It plans to double down on AI innovation and deepen its global presence by venturing into the UK and Australia.
Dazz raises $50M in funding for their automated cloud security recommendation.
Alphabet pours $5B into Waymo Cruise scraps the Origin and Elon’s bet on autonomy.
Cybersecurity firm Wiz calls off $23 billion deal with Google as it plans to grow on its own and focus on potential IPO.
OneStream IPO Raises $490 Million Priced Above Range and rang the opening bell on July 24, priced at $20 per share.
Applied Intuition closes $300M secondary four months after raising $250M.
Was this forwarded to you?
That’s it for this week. I hope it was insightful. As always, let me know what you think and if you have any questions. Cheers!
🌜 Loryn and Nicole from Dark Mode Digest
How'd you like today's digest? |