You can use serverless endpoints on runpod or similar services the delay is a bit high for the first inference but for concurrent requests it's fine with a reas

You can use serverless endpoints on runpod or similar services the delay is a bit high for the first inference but for concurrent requests it's fine with a reasonable keep alive.