You can always distill VLMs into much smaller / faster models that’s specific to your domain or use-case. What’s the use-case and what kind of latency do you r

You can always distill VLMs into much smaller / faster models that’s specific to your domain or use-case. What’s the use-case and what kind of latency do you require?