Open vs closed source

🔒 Should you use an open or a closed AI solution?

This is one of the most common questions we get at Mantis at the start of a project as a lot of clients are wondering what’s the best choice. In one word, this comes down to control 🕹️  In particular, think about the following requirements:

🚥 Availability: should the AI be always available?

✋ Rate limit and response time: how many times and how quickly do you need a response from the AI model?

💬 Customisation: how should the AI respond?

🌐 Data processing: where is data processed and stored?

💰 Cost: what is the cost per request that works for your business?

Note how I did not include performance in the requirements since open source solutions are to a large extent on par with closed source ones 💪

The more control you need, the more the balance tilts towards open source; whereas the more an off-the-shelf, just-works solution you want, the more the balance tilts towards closed source or some third-party provider.

A short answer thus would be to use closed source, or third party, for prototyping and open source, or hosted in your infrastructure, for production ✨

🔗 Read more on this topic in this blog post we have written https://medium.com/mantisnlp/how-were-thinking-about-generative-ai-proprietary-vs-open-source-3c8aff23a82b