New in August: Deploy, deploy, deploy

What we’re reading and writing

Model deployment

Machine learning models lounge around in Jupyter notebooks, but to drive business value, they need to venture into the real world. Our new model deployment page walks you through quickly deploying a model in any supported framework or using Truss to gain greater control over the deployment process. Every example is accompanied by a runnable Google Colab notebook that trains, deploys, and invokes an ML model in a supported framework.

The new model deployment page

Thanks to Truss, you can now access secret values directly in models run on Baseten. Start by adding secrets to your Truss, then add the secrets and values in your Baseten workspace. When your model is deployed, it will securely read the values and use them for accessing things like AWS resources, APIs, and databases.

We’ve also introduced a number of quality-of-life improvements to the model deployment process, including a new pre-trained models exploration page, email notifications for model deployment status, and a model version field in model blocks. And we’ll keep chronicling new features in our changelog, published weekly.

Plans and pricing

Now that teams and developers around the world are getting value from Baseten, we introduced the first version of our starter and business plans to give our power users the resources they need to scale.

We’re excited about the features available to paid workspaces, such as:

  • A draft environment for testing changes before deployment

  • More model resources, including GPUs, with usage-based pricing

  • Coming soon, an integration for backing up your Baseten applications to GitHub

When you're ready to upgrade, let us know and we'll work with you to define a plan that fits your needs.

New Baseten Pricing

Inside Baseten

In team news, two talented additions: Abu, a software engineer, and Jesse, a developer advocate. We might be looking for you, too! We’re hiring for a lead ML engineer to build the infrastructure and services for our users to deploy, serve, and monitor their ML models.

That’s it for now, but this newsletter will be back up when September ends.

Thanks all,

The team at Baseten