MAI-1 - and in general LLM names are looking more and more like names of…

MAI-1 (https://arstechnica.com/information-technology/2024/05/microsoft-developing-mai-1-language-model-that-may-compete-with-openai-report/)
- and in general LLM names are looking more and more like names of spacecraft. Voyager, Soyuz, TKS and so on.
In short, Microsoft hired a former DeepMind leader and now, obviously, is training its own foundation model with 500B parameters, most likely a release at their Build conf (https://www.theverge.com/2024/2/14/24073220/microsoft-build-event-schedule-may-21)
on May 21. Well, we’ll see how they’re going to compete with their own startup. Reminds me of the DeepMind & Google setup.
And in general
here’s a useful resource (https://huggingface.co/spaces/ArtificialAnalysis/LLM-Performance-Leaderboard)
for comparing price/speed of available model providers. Hope Microsoft won’t make it as expensive as OpenAI.