Simplifying billing: Update to “per-second” based billing system for serverless APIs

To help simplify the billing process, we are now moving to a per-second based billing throughout the platform from the 1st of May, 2024

Simplifying billing: Update to “per-second” based billing system for serverless APIs

TL;DR: We are moving to a simpler, per-second-based billing starting from the 1st of May, 2024.

Why we are moving to a new billing system?

We are constantly trying to improve the user experience and make our platform more developer-friendly. To help simplify the billing process, we are now moving to a per-second based billing throughout the platform. We have shared the details below on how our new billing will work. But before that, I wanted to share why we are moving to this new system.

  1. Introduction of fine-tuning service.  We introduced fine-tuning back in December of last year. We had to create a whole new billing system that was based on per-second billing. This was needed since each job could be anywhere between a few minutes to a few hours depending on your data size and parameters.
  2. Introduction of a new class of image and video models.  We wanted to introduce a new set of models that have a different set of parameters compared to stable diffusion. For Stable diffusion, steps and schedulers were key parameters that determined the price. But all the models on our platform that were not SD-based required a different framework to bill (eg. ESRGAN, FaceSwap, etc).

    Hence we decided to move everything to “per second” pricing.

New Billing Details

You can find the details below for billing updates for inferences, fine-tuning, our newly launched workflows tool - PixelFlow, and changes in the billing cycle.

Billing for Inferences

Inferences will be billed on the number of seconds it takes. Each model runs on different GPUs and the per-second pricing for each model can be found on each Model Page.

You should see the usage as it is shown today, with the only change being that we will be adding a new column called “average duration” to show you the average time each model takes to execute.

Billing for Fine-tuning

Fine-tuning will be charged at a flat rate of $0.001 per second.  Fine-tuning a model can take anywhere from a few minutes to a few hours.  To learn more about fine-tuning, take a look at our Docs.

PixelFlow Subscription

We are excited to announce the release of our new Workflow Platform, PixelFlow. Read our announcement here.  To start, all new users will be granted free credits and access to explore Workflow Templates, create new workflows of their own, and generate outputs.  For users who wish to subscribe, we will be offering a paid “Pro” tier for $49 USD per month.  For the first 1000 users, we are offering a discount, down to $29 USD introductory rate.  Pro Tier users will get “unlimited” access to PixelFlow GUI and Model APIs. You also be able to use will also get limited access to Workflow APIs. Any workflow that a user likes can be converted to an API for easy integration into any application or service.

Please Note about "unlimited":  PixelFlow users will be charged based on the number of seconds taken to process the sum-of-the-parts. Pro Tier users will be metered and access will have some limit to ensure there is no abuse of the platform.  If a Pro Tier user reaches this upper limit, they will be notified and directed upgrade to the Enterprise Tier.

Changes in the Billing Cycle

We are also making one more minor change. For our current Post-Paid customers, we are moving from twice-per-month billing to once-per-month billing.

Our Commitment to You

We understand that changes to billing systems can raise questions, so we want to assure you that we're committed to minimizing any significant impact on your monthly bill. Our dedicated team is standing by to assist with pricing adjustments if needed, ensuring a seamless transition for you. If you have any questions or concerns about the new billing system, please don't hesitate to reach out us here. We're here to help and support you every step of the way.  

We’re excited to work with developers who are creating the next generation of Generative AI applications.  Please join our discord community to connect with us.