The New King of AI 👑

Over this issue, we will discuss Claude 3, Inflection 2.5, TripoSR, top AI Tools and news from this week. Let's jump in!

March 11, 2024

You read that right. After an year, GPT-4 is no longer the best AI Model in the World. And a lot of users are now switching. For context, we have 2 models better than GPT-4. You’ll know which 2 models are better. For quick overview: today we cover Claude 3, TripoSR, Inflection-2.5.

I think you got the hint. BTW, I’ve also got a gift for you near the end. You’re not gonna regret it.

Let’s get rolling.

Stuff you should know

Claude-3 👑

Anthropic has introduced the Claude 3 model family, a new generation of AI that promises to redefine the standards of intelligence in the industry. The family includes three models: Claude 3 Haiku, Claude 3 Sonnet, and Claude 3 Opus, each offering different levels of performance to cater to various applications. Claude 3 boasts near-human comprehension, strong vision capabilities, and improved accuracy, with fewer refusals and a more nuanced understanding of requests.

Stats and Bullets

  • Claude 3 Opus outperforms peers on benchmarks like MMLU, GPQA, GSM8K

  • Near-perfect recall with 99% accuracy on the 'Needle In A Haystack' evaluation

  • Claude 3 models can process visual formats including photos, charts, and diagrams

  • Claude 3 Opus has surpassed GPT-4 on 10 AI benchmarks

  • Claude 3 models are available in 159 countries, with Opus and Sonnet already accessible

  • 200K Context Window!

This is right now the best Multi-modal model available, for day to day use cases.

Pi, Inflection-2.5

Inflection's Pi AI, powered by the new Inflection-2.5 LLM, is distinguished by its dual focus on delivering both factual information and emotional support. It aims to be a personal AI capable of engaging in a broad spectrum of conversations, from discussing current events to providing companionship and empathy. With the Inflection-2.5 update, Pi has significantly improved in coding and mathematics, and it now features real-time web search capabilities. This enhancement has made Pi a formidable competitor to other leading AI models like OpenAI's GPT-4, offering nearly the same level of performance while requiring considerably less computational power for training.

Stats and Bullets:

  • Inflection-2.5 achieves over 94% of the average performance of GPT-4.

  • Pi has an active user base of 1 million daily and 6 million monthly.

  • Users have exchanged over four billion messages with Pi.

  • The average conversation with Pi lasts 33 minutes, with 10% extending over an hour.

  • Around 60% of users return to Pi the following week.

  • Inflection-2.5 was trained using only 40% of the computational resources compared to GPT-4.

TripoSR

TripoSR is a groundbreaking AI model developed by Stability AI and Tripo AI, designed to generate high-quality 3D models from single images in under a second. This model stands out for its speed, efficiency, and accessibility, making it a valuable tool for professionals in entertainment, gaming, industrial design, and architecture. With its open-source availability and low hardware requirements, TripoSR is poised to revolutionize the way we create and visualize 3D objects.

Stats and Bullets:

  • Speed and Efficiency: Generates draft-quality 3D outputs in about 0.5 seconds on an Nvidia A100, significantly outperforming other models like OpenLRM.

  • Accessibility: Can run under low inference budgets and without a GPU, making it accessible to a wide range of users.

  • Open Source: Model weights and source code are available under the MIT license for commercial, personal, and research use.

  • Technical Improvements: Includes channel number optimization, mask supervision, and a more efficient crop rendering strategy over the base LRM model.

  • Training Data: Utilizes a carefully curated subset of the Objaverse dataset, enhancing the model's ability to generalize.

Watch it in action here.

Stable Diffusion 3

Stable Diffusion 3 is the latest iteration of the text-to-image AI model developed by Stability AI, designed to generate high-quality images from textual prompts. It introduces significant improvements in image quality, multi-subject prompt handling, and text representation within images. The model suite ranges from 800 million to 8 billion parameters, catering to various user needs and hardware capabilities.

Stats and bullets

  • Over 10 million users across all channels.

  • More than 12 billion images created using Stable Diffusion.

  • Model sizes range from 800 million to 8 billion parameters.

  • Improved performance in multi-subject prompts, image quality, and spelling abilities.

  • Open-source and accessible under a permissive license.

Compared each of the prompts with DALLE-3 and i think open source has won!

Top Products launched this week

  • LTX Studio -  Your all-in-one AI filmmaking platform. From ideation to production, control every aspect of your story for elevated outcomes that realize your vision.

  • SPUN - Simplify visas & permits for Indonesia with our online platform. Sign up for free, choose from 100+ permits, submit documents online, and get permits on time.

  • humaan.ai - Humaan.ai: Access curated AI models for easy integration. Build intelligent apps without deep AI knowledge.

  • AI Podcast Transcription - AI audio-to-text transcription service: Speaker detection, easy editing, multiple download formats, and shareable timestamps for each episode.

  • RipeOrNot AI - For Avocados - Avocados are great. But only the ripe ones. No more guesswork. Use AI to help you determine whether your avocado is ripe or not — no more feeling up the avocado and popping the stem. Just click a picture, and you're done!

  • Anytalk - Anytalk.ai: Translate video and audio streams instantly. Join multilingual video conferences with minimal delay. Exciting updates ahead: Soon, effortlessly speak any language with our upcoming app

  • Wondercraft - Wondercraft: Create studio-quality audio for podcasts, audiobooks, ads, and more. Effortlessly translate content for a global audience, all in one place.

  • Claude 3 - The three state-of-the-art models—Claude 3 Opus, Claude 3 Sonnet, and Claude 3 Haiku—set new industry benchmarks across reasoning, math, coding, multilingual understanding, and vision.

  • Lummi - Explore Lummi's collection of 13,000+ AI-crafted, highly curated images from creators across the globe. Bold portraits, tranquil landscapes, and more. Effortless content creation with our Figma integration.

  • Vmaker AI Video Editor - Vmaker AI: Turn raw videos into impressive content in minutes. Automatically add b-rolls, music, transitions, subtitles, and more. Easily create short-form videos, teasers, and highlights.

  • Athina AI - Athina helps developers monitor and evaluate LLMs in production. Get complete visibility into RAG pipeline and 40+ preset eval metrics to detect hallucinations and measure performance.

  • DOO - DOO makes hiring AI mates easy. Our mates are expertly trained for seamless collaboration, boosting productivity in your workplace

  • Distillery 2.0 - Open-source AI image generator offering deep control for professional Generative Art. Master Generative AI for your use cases with our free course

  • Picurious AI - Picurious leverages GPT-4V, turning photos into learning moments. Get instant answers, insights, and inquisitive questions about your pictures, and easily identify any object, plant, or text.

  • Travelicious (Beta) - Travelicious is your next-generation AI travel planner focused exclusively on city adventures. Experience cities like never before with tailored recommendations for dining, exploring, and enjoying.

Handpicked video

Another OpenAI Drama, don’t forget popcorn!

A Gift 🎁

This week only, you can get 90% OFF on any of the AI Courses below.

This week in AI

Hope this was a helpful issue. That’s going to be all.