AI with top frontier models is like restaurant food — you have to visit OpenAI, Google or Microsoft to consume it. But that is slowly changing.
Top AI companies are making takeout versions of their proprietary frontier large language models, which users can run in their own data centers.
Companies can also use custom inference models, with or without internet connections.
Google is customizing versions of its Gemini LLM that companies can install on servers with GPUs in their data centers. That’s a big change from the past when Gemini was…
