Published on

Changelist: May 2024

Written by
Codeium Team

Stack Overflow just came out with insights on how developers use and view AI assistance tools, and Codeium was at the top right when it comes to productivity and satisfaction. We are continuing to work hard to drive even more value to our community. While we appreciate all of our users, a core thesis to Codeium is to be valuable even in the most complex enterprise environments, so it was also great to see Codeium gaining traction amongst professional developers.

StackOverflow pulse survey results.

Hybrid Deployment

The big announcement this month is a first-of-its-kind enterprise deployment model for an AI code assistant tool, which we are dubbing Hybrid. As some context, Codeium has both a data layer (precomputed data structures for context awareness, attribution and audit logs, etc) and a compute layer (GPUs for inference) component. Until now, we have had two main deployment methods:

  • Fully self-hosted: where both the compute layer is run and the data layer is stored within a customer’s tenant
  • Fully SaaS, where both happen on Codeium’s hosted servers.

The comparative benefit of self-hosted is additional security around code IP since retained code snippets and any derived data structures are still only available within the company’s firewall, while the comparative benefits of SaaS are not needing to manage and pay for expensive GPU-enabled hardware as well as getting earlier access to new features and capabilities. Hybrid keeps the data layer on the customer’s side while hosting the compute layer on Codeium’s servers, which gives the best of both worlds:

  • No GPU cost and management
  • Full data security because Codeium’s servers come with SOC2 Type2 certified zero data retention guarantees
  • Ability to use features that depend on retained derived data like remote indexing and audit/attribution logs
  • Early access to new capabilities

We believe that Hybrid is the right deployment model for the vast majority of companies, and are excited to be the first company to offer such a deployment. Please reach out on our contact page if you have interest, and we will be publishing more content around the Hybrid offering shortly.

Chat and Command get 4x context length

In our ever-ending quest to get every developer the best product, we have increased the context lengths for all of our Chat and Command models by 4x, for both SaaS and self-hosted customers. This allows for reasoning over larger quantities of relevant code (thanks to our context awareness system) as well as longer answers for more open-ended chat questions.

Startup Starter Kit

Are you getting your own startup off the ground or know someone who is? We teamed up with our friends at Graphite, Statsig, Digital Ocean, Stytch, and Temporal to get you deals to start off on the right foot with the right tooling. You can sign up for Codeium’s offer, 6 months free of Teams, here.


Some content on Codeium from around the Internet:

Also, we hit 1 million downloads on our Visual Studio Code extension alone last month! Excited for the value that Codeium will be able to provide the next million(s).

Explore more

May 8, 2024product4 min

Codeium updates from April 2024!

Apr 9, 2024product2 min

Codeium updates from March 2024!

Mar 7, 2024product3 min

Codeium updates from February 2024!

Stay up to date on the latest Codeium & AI dev tool news

Please enter a valid email address