Mixtral 8x7B: 24 Hours Challenge

At this event, you are challenged to create your own product with the Mixtral 8x7B LLM that overperformed GPT 3.5, and Llama 2 70B on most of the benchmarks.

• 🕒 24 hours to complete the challenge.
• 💡 Implement any idea that uses Mixtral 8x7B in its core.
• 🤝 Take part alone or form a team with other participants
• 🚀 The final project can be used for your portfolio, and the startup application to the Slingshot accelerator

Enroll now!

Check the apps build with Mistral at lablab.ai

Mistral and Chroma Vector Database RAG Application

Improving Mistral LLM with Chroma Database to build a better chat and document-saving tool.

Socratic-AI

Socratic-AI: An easy-to-use chatbot that helps you learn and think deeply using smart conversations.

Customer Support Chatbot for Twitter (X)

Mistral-7B is a revolutionary Twitter customer support chatbot that remembers conversations and excels in handling a wide range of issues.

Register to start the challenge

Test out your skills prompting and building actual projects with LLMs.

Enroll now!

Here’s what is that exiting about Mixtral 8x7B

Mixtral 8x7B is an advanced, efficient computer model good at understanding and generating text in five languages. It's fast, performs well in coding tasks, and is easy to adjust for specific uses. In most benchmarks, Mixtral matches or outperforms Llama 2 70B, as well as GPT3.5 models.

Mixtral's abilities include:

It works well with up to 32,000 tokens.
It can use English, French, Italian, German, and Spanish.
It's really good at code generation.
Mixtral can be finetuned into an instruction-following model that achieves a score of 8.3 on MT-Bench.

The useful resources:

Mixtral’s model information
Download the model
Register on Mistral AI to access more generation models and embeddings

Guides from creators:

Mixtral On Your Computer from Martin Thissen
Mixtral 8x7B - Mixture of Experts dominates other models by Matthew Berman

This Hackathon is for you if:

You are making your first steps with LLMs: Mixtral will allow you to test and learn the basics of working with LLMs. Use the tutorials mentioned above, research yourself, and, most importantly, try to practice!
You have an idea you want to test: Even with no technical skills, the hackathon format allows you to connect with other members of the lablab.ai AI community to build together! Work on your idea concept, create a presentation, or even learn coding yourself, supporting your team members!
You are an experienced developer: DIt’s a great opportunity for you to try out the newly released model, compare its capabilities and output with other LLMs you’ve worked with. The challenge will allow you to make a “promise” and dedicate yourself to exercising with Mixtral!