MythoTech AI News: 5/16/2024

A new champion has been crowned... 🥇

GPT-4o is the biggest news in AI this week, deservedly so. Today it hit the Chatbot Arena leaderboard like a storm with the most impressive debut ever seen - a whopping ❗39❗ points above 2nd place (GPT-4 Turbo).

This model, more than anything else I've seen, really encapsulates the insane growth we're continuing to see in the LLM space. Half the price for the API. Free for everyone to try. Twice the speed. The highest benchmarks. Natively multimodal.

And of course, casually dropped on the eve of Google I/O. 😏

Google announced some interesting improvements for Gemini, and the model continues to score decently well on the Chatbot Arena leaderboard - but I can't really recommend it to anyone for business use while it lingers in "private preview" until sometime next month.

Speaking of my recommendations, check out the latest scores and data points for all the publicly available APIs at the MythoTech LLM API Factsheet.

And some rapid-fire analysis of where the models stand as of today for business use -

Best overall: GPT-4o
Fastest high end model: Llama 3 70b Instruct (running on Groq)
Best writer: Claude Opus
Best for coding: GPT-4o
Best open-source: Llama 3 70b Instruct
Best value (simple tasks): Llama 3 70b Instruct
Best value (complex tasks): GPT-4o

---

This content was originally posted on LinkedIn, and we invite you to join the conversation there!