Mistral NeMo, GPT-4o mini, AI-powered platform to create controlled videos, SmolLM, Anthology Fund and more

Hi. Welcome to this week's AI Brews for a concise roundup of the week's major developments in AI.

In today’s issue (Issue #71 ):

1. AI Pulse: Weekly News & Insights at a Glance

2. AI Toolbox: Product Picks of the Week

---

🗞️🗞️ AI Pulse: Weekly News & Insights at a Glance

🔥 News

1. Mistral AI and NVIDIA released Mistral NeMo, a new state-of-the-art 12B model with 128k context length. The model is designed for global, multilingual applications and is released under the Apache 2.0 license. It is trained on function calling and is particularly strong in English, French, German, Spanish, Italian, Portuguese, Chinese, Japanese, Korean, Arabic, and Hindi [Details].

2. Mistral AI released two small models under the Apache 2 license: Mathstral 7B and Codestral Mamba 7B. Mathstral 7B is a model specializing in mathematical and scientific tasks and is based on Mistral 7B . Codestral Mamba is one of the first open source models with a Mamba 2 architecture.

3. Hugging Face research team introduced SmolLM, a family of state-of-the-art small models with 135M, 360M, and 1.7B parameters, trained on a new high-quality dataset that outperform MobileLLM, Phi1.5 and Qwen2 small models [Details].

4. Groq released two new open-source models specifically designed for tool use: Llama-3-Groq-70B-Tool-Use and Llama-3-Groq-8B-Tool-Use, built with Meta Llama-3. Llama-3-Groq-70B-Tool-Use is the highest performing model on the Berkeley Function Calling Leaderboard (BFCL), outperforming all other open source as well as proprietary ones like Claude Sonnet 3.5, GPT-4 Turbo, GPT-4o and Gemini 1.5 Pro. Both models are now available on GroqCloud™ Developer Hub and on Hugging Face [Details].

5. OpenAI announced GPT-4o mini - a cost-efficient small model that surpasses GPT-3.5 Turbo and other small models on academic benchmarks across both textual intelligence and multimodal reasoning, while being 60% cheaper than GPT-3.5 Turbo [Details].

6. DeepSeek has open-sourced DeepSeek-V2-0628 checkpoint, the top open-source model on the LMSYS Chatbot Arena Leaderboard [Hugging Face].

7. Anthropic and Menlo Ventures launched Anthology Fund, a $100 million fund for AI startups [Details].

8. Andrej Karpathy, founding member OpenAI/former Director AI at Tesla, is starting an AI+Education company called Eureka Labs that will incorporate AI teaching assistants in its platform. The first product will be an undergraduate-level AI course, LLM101n. [Details].

9. Haiper, the AI video startup founded by former Google Deepmind researchers launched Haiper 1.5, a new visual foundation model that allows users to generate 8-second-long clips from text, image and video prompts [Details].

10. Hugging Face released Docmatix, a huge dataset to enhance vision-language models for document understanding [Details].

11. OpenAI launched a new Enterprise Compliance API, admin tools and GPT controls to support compliance programs, data security, and user access at scale for ChatGPT Enterprise customers [Details].

12. Google is open-sourcing Project Oscar, a project to improve open-source software development by creating automated help, or “agents,” for open-source maintenance [Details].

13. Anthropic has doubled the max output token limit for Claude 3.5 Sonnet from 4096 to 8192 in the Anthropic API [Details].

14. Snowflake released an updated version of snowflake-arctic-embed-m designed to improve embedding vector compressibility. snowflake-arctic-embed-m-v1.5, is capable of producing highly compressible embedding vectors that preserve quality even when squished as small as 128 bytes per vector [Hugging Face]

15. Exa raises $17M from Lightspeed, Nvidia, Y Combinator to build a Google for AIs [Details]

16. Google is rolling out its new Vids productivity app in Workspace Labs . Vids allows users to drop docs, slides, voiceovers, and video recordings into a timeline to create a presentation video to share with coworkers [Details].

17. Microsoft introduced SpreadsheetLLM, an efficient encoding method to optimize LLMs' understanding and reasoning capability on spreadsheets [Paper].

18. Applications are open for batch 4 of AI Grant [Details].

19. The full and final text of the EU AI Act, the European Union’s risk-based regulation for applications of artificial intelligence, has been published in the bloc’s Official Journal. The framework puts different obligations on AI developers, depending on use cases and perceived risk [Details].

20. YouTube Music is testing an AI-generated radio feature and adding a song recognition tool [Details].

21. Anthropic’s Claude Android app is available on Google Play.

22. Rufus, Amazon’s generative AI-powered conversational shopping assistant is now available to all U.S. customers in the Amazon Shopping app [Details].

23. Salesforce debuts Einstein Service Agent, a new AI Agent for customer self-service [Details].

24. Meta will withhold its next multimodal AI model — and future ones — from customers in the European Union because of what it says is a lack of clarity from regulators there [Details].

🔦 Weekly Spotlight

1. Using Claude 3.5 Sonnet and Supabase to build an e-commerce app with AI powered related-items search [Link]

2. The Data Executive’s Guide to Effective AI by Snowflake - Best Practices from Data Executives for an AI Transformation Journey [Link].

3. PraisonAI: Combines AutoGen and CrewAI or similar frameworks into a low-code solution for building and managing multi-agent LLM systems [Link].

4. Introducing the Next Generation of AutoGPT [Link].

5. The Future of Knowledge Assistants - by LlamaIndex founder & CEO Jerry Liu [Link].

6. How a small business owner saved $20,000 with Gemini Flash [Link]

7. Qwen2 Technical report [Link]

---

🔍 🛠️ AI Toolbox: Product Picks of the Week

1. Morphic Studio: an AI-driven controlled-video creation platform designed for filmmakers, animators, and creators. You can join the waitlist here

2. Nailedit: Compare multiple LLMs side by side with one prompt

3. Buildbox 4 AI: Design and build games with generative AI.

4. Microsoft Designer: Create images, stickers, greeting cards, collages, custom brand kits etc. with AI. Designer is integrated across Word and PowerPoint. Designer is now also available as a free mobile app (iOS and Android) along with the web version.
   

Last week’s issue

You can support my work via BuyMeaCoffee.

Thanks for reading and have a nice weekend! 🎉 Mariam.