Skip to content

Comment on AWS AI Stack – Ready-to-Deploy Serverless AI App on AWS and Bedrockparent

Comments

They use GPUs under the hood for inference/fine-tuning and charge by token. Fireworks will even let you deploy a Lora serverless at the same pricing as base model.

But not aware of any “lambda”-like serverless for any old CUDA workload. Given loading times, it wouldn’t really make sense. Something like CloudRun or KNative for GPUs would be cool.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.