> ended up enabling groq
For those reading fast, this isn't a reference to SpaceX's Grok, this is Groq.com - with its custom inference chip, and offerings like https://groq.com/blog/introducing-llama-3-groq-tool-use-mode... and https://console.groq.com/landing/llama-api
Really liked Groq due to its speed but it seems like after Nvidia bought it it has been discontinued...