Skip to main content

Public Model

  • Updated

Public Model

Model Strength

Basic / Advanced

Speed or Strength

Do you prefer a lighter, faster computational load or the raw power and accuracy of a massive AI model - you choose:

  • Speed Runs fast and light using fewer data points - Recommended for machines with < 4GB RAM (81 million parameters)

  • Balanced Get the best of both with a balanced approach - Recommended for machines with 4-32GB RAM (124 million parameters)

  • Strength Higher accuracy requiring more computational resources - Recommended for machines with > 32GB RAM (355 million parameters)

Use specialized models

Tabnine keeps adding more awesome value for your Team! If your team creates code using Dart, Julia, HCL, Ruby, Rust, C++, Python, Go, or Java, your Tabnine AI assistant can run one of our new specialized models tailored to your team’s specific dev stack.  

 

Model Serving

cloud vs. local

The cloud version eliminates 90% of CPU consumption and 70% of memory consumption.
If you choose to use Tabnine Cloud, Tabnine will send blocks of code from your edited files to our server, allowing us to provide deep completion suggestions. These blocks of code will never be stored - they are used to calculate predictions and then immediately discarded.

The main difference is that the local model works offline and the cloud model is not. But there is no difference between the models in terms of the strength of the completions.
If you have a stable internet connection - you should use the cloud version.

 

Related articles