• unwind ai
  • Posts
  • New Opensource LLM beats Mixtral and Llama-2

New Opensource LLM beats Mixtral and Llama-2

PLUS: xAI's Grok for all premium subscribers, New AI pendent, Train LLM with just a prompt

Today’s top AI Highlights:

  1. Elon Musk to enable xAI’s chatbot Grok for all premium subscribers

  2. DBRX by Databricks beats Mixtral, Llama-2 and GPT-3.5

  3. Airtable enables AI features to streamline business workflows

  4. Claude LLM trainer lets you train LLM with just a text prompt

& so much more!

Read time: 3 mins

Latest Developments 🌍

Grok for All Premium Subscribers 🤖

Elon Musk has announced that xAI’s chatbot Grok will be available to all premium subscribers on X later this week. X Premium has three tiers: Basic, Premium and Premium +, with more features available to high-paying subscribers. Currently, Grok is available only to Premium+ subscribers.

Musk is also enticing users to upgrade to Premium by introducing an ‘unhinged fun mode’ to Grok. Grok operates in two modes: the default Fun Mode, which provides witty and humorous responses, and Normal Mode, which delivers more precise answers, similar to other chatbots.

🤔 Why do you think Elon Musk is making Grok more accessible? Reply in the comments below!

New Opensource LLM Beating Mixtral and Llama 2 💪

Databricks has opensourced their latest offering, DBRX, a general purpose LLM leveraging Mixture-of-experts (MoE) architecture, that outperforms all established open source models including Llama-2 70B, Mixtral 8x7B, and Grok-1 on all industry benchmarks. Not just the performance, its speed is impressive with inference up to 2x faster than Llama 2 70B. The model lets you have your cake and eat it too in terms of speed (tokens/second) vs performance (quality).

Key Highlights:

1. Beats opensource LLMs: Outperforms Mixtral MoE, Llama-2 70B and Grok-1 on language understanding , Programming, and Math.

2. Beats closed source LLMs: Surpasses GPT-3.5, and is competitive with Gemini 1.0 Pro. Equally good for specialized models like CodeLLaMA-70B unlike other foundational LLMs.

3. Faster Inference and Less Memory: Inference is up to 2x faster than Llama-2 70B, and DBRX is about 40% of the size of Grok-1 in terms of both total and active parameter counts.

4. Interesting Architectural Choices: DBRX features 16 experts, from which it selects 4 based on the input. This contrasts with other models that might have fewer experts and thus, fewer combinations of experts to choose from. It uses the GPT-4 tokenizer, which is part of the Tiktoken repository.

5. Opensource: Base model (DBRX Base) and the finetuned model (DBRX Instruct) weights are made available under an open license on Hugging Face.

AI Pendant that Acts as Your Second Brain 💎

As we’ve moved through the evolution from mainframe computers to PCs, laptops, smartphones, and now smartwatches, our devices have increasingly become integral parts of our lives. But now we want devices that are not just tools but personal companions, seamlessly integrated into our daily routines. Entering this space is a new, pendant-sized AI wearable that turns your experiences into a personal AI assistant.

Tone is an AI-enabled pendant that is as small as a quarter (approximately 20% smaller than its closest competitor). Priced at $299, it features a solid aluminum body and maintains an all-day battery life.

Tone acts as a co-pilot for our daily lives by providing perfect memory recall and acting as a personal assistant. It’s designed to automate trivial tasks, from managing our calendar to reminding important personal moments, streamlining day-to-day activities, and making sure we’re always on top of our game.

Airtable Rolls out New AI Features 🥁

Airtable is rolling out new AI capabilities in Airtable AI to cater to the expansive needs of enterprises. With hundreds of AI tools launching every year for several usecases, it becomes complicated for a business to adopt AI across multiple functions, manage tool sprawls, and streamline workflows. Airtable’s new features emphasize practical AI applications and transform digital processes using AI at scale.

  • AI Features:

    • Automatic Summarization: Condense large volumes of information quickly.

    • Data Categorization: Organize data at scale efficiently.

    • Content Generation: Create customized content for various communication platforms.

    • Instant Translation: Translate texts between languages in real-time.

    • Task Routing: Automatically direct tasks to appropriate team members.

  • AI Model Diversity:

    • Selection Freedom: Choose from multiple leading AI models, including:

      • OpenAI GPT-3.5 and GPT-4

      • Anthropic's Claude and Claude 2

    • Platform: Models are accessible through Amazon Bedrock.

  • Accessibility:

    • Service Type: Paid, opt-in service.

    • Flexibility: Enable AI features for specific workspaces or across the company.

    • Management: Control through an Admin Panel.

AI-UseCase-Product-3-Generate-1650x928@2x.webp

Tools of the Trade ⚒️

  1. Keak: An online platform designed to automate and manage conversion rate optimization for websites, where AI generates variations for website content to test, and previous tests are used to fine-tune your personalized website model, ensuring it becomes more adept at understanding your audience’s preferences with each cycle.

  1. Surya: An OCR toolkit designed for document analysis, supporting over 90 languages. It excels in accurate OCR, line-level text detection, and layout analysis for various document types, leveraging PyTorch and offering an interactive application for testing.

  2. Claude LLM Trainer: The simplest way to train a task-specific LLM. Simply input a description of your task, and the system will generate a dataset from scratch, parse it into the right format, and fine-tune a Llama 2 or GPT-3.5 model for you.

  3. Coverr’s AI Workflows: Master the art of AI video generation by discovering AI-generated footage, the tools used for creation, and prompts used by a community of AI video experts.

😍 Enjoying so far, share with your friends!

Hot Takes 🔥

  1. Reasoning might be the big LLM unlock to unleash another 10x expansion of viable use cases. ~Logan.GPT

  2. We don't need GPT-5 to write better copy, chat with us in a more natural way, etc. The tech that's out there today can solve most of the problems we're already thinking about. Instead, GPT-5 is going to open up massive new use-cases that we can't even imagine doing with GPT-4. ~Matt Shumer

Meme of the Day 🤡

That’s all for today! See you tomorrow with more such AI-filled content.

Real-time AI Updates 🚨

⚡️ Follow me on Twitter @Saboo_Shubham for lightning-fast AI updates and never miss what’s trending!

PS: I curate this AI newsletter every day for FREE, your support is what keeps me going. If you find value in what you read, share it with your friends by clicking the share button below!

Reply

or to participate.