A new open-source LLM to consider. A NY Times article: How Chinese A.I. Start-Up DeepSeek Is Competing With Silicon Valley Giants
I think it deserves a mention because of this from the NY Times article: "DeepSeek is run by a quantitative stock trading firm called High Flyer. "
So potentially designed from the outset for quantitative finance, open source, more computational efficient that ChatGPT, less code, and on-par with ChatGPT for most metrics.
Mixture-of-Experts (MoE) Architecture:
- Like Grok which I find to be super-fast, DeepSeek employs a Mixture-of-Experts model, a cutting-edge approach that activates only a subset of its parameters for a given task. This makes it computationally efficient while maintaining high performance.
Fortune. Use the reader with Safari to bypass the paywall.
" It’s David versus Goliath in the world of AI.
On one side: The U.S. and its new Stargate Project, a massive $500 billion pledge with backing from the biggest names in tech to build data centers across the U.S. On the other, a scrappy Chinese startup that’s released an impressive open-source model trained on a shoestring budget."