Our v2 API (including new Custom Training and Visual search features) is graduating from developer preview to general availability. Among the many new features in our v2 API, usage-based pricing is one of the biggest changes we've made to help developers validate their ideas, go to market faster, and scale their apps by paying only for what they use. Learn more about our new pricing and how you can take advantage of it!
One of our goals at Clarifai is to provide developers with the tools they need to easily build cutting-edge, sophisticated artificial intelligence into their apps. On the technical side, our new v2 API takes all the grunt work out of building, training, and using AI (see the docs!). On the business side, our new API pricing is designed to allow developers like you test and validate your ideas and pay only for what you use.
So, when you're hacking away at building the next Snapchat or creating a proof-of-concept for a project, you'll be able to use Clarifai to power your image recognition at little to no cost. And, when your app blows up in the iTunes store and gets a billion users, we'll be right there to help you scale.
So, without further ado, read on for more information about how our new usage-based pricing works!
V2 Pricing Tutorial
Clarifai’s v2 API charges you each time you perform an API operation, so you only pay for what you use. You will be charged each month, on the anniversary of your account creation, and can view your invoices and usage history in the account section on our developer website.
You can get started with the Clarifai API for free! You can perform up to 5,000 operations, index 10,000 inputs, and maintain 10 custom concepts for free ... forever. This works out great for low-volume applications, side projects, proof of concepts, or the ever-important “get to know us” phase of our relationship. Once your app takes off, you can head over to our developer website and add your credit card to upgrade to our paid plan.
There are four types of actions that we charge for:
- Search and predict from an input;
- Index an input;
- Create a custom model; and
- Use a custom model.
For more clarification on how these actions and terms are defined, please read their definitions in our developer documentation glossary!
Search and Predict
Search and Predict operations are the primary ways to extract information from Clarifai.
When you pass an input (AKA an image) to `models.outputs`, you are asking Clarifai to analyze it and report back with what it knows. Alternatively, you can pass an input to `post.inputs` to both analyze it and index it for future use within an application. Both operations are charged as one Standard Operation at $1.20 per thousand inputs.
For inputs that you add to an application, we keep them around and use them as the potential result set for any searches that your perform, and as training data for custom models. We charge a monthly fee of $0.80 per thousand Indexed Inputs to maintain and optimize them for you. If you no longer want an input to be included in search results and custom model training, you can delete them at any time for free.
Once you have a set of indexed images, you can perform all types of searches over them to find the ones you need. Performing a new search is counted as a Standard Operation. You can also recall previous results by search ID for free, so duplicate searches don’t count against you.
Train your own model
Besides search, you can also use Indexed Inputs to create custom models that meet your unique specifications. To do this, you’ll use a couple of Standard Operations to create and train the model, which is defined by the number of Custom Concepts it contains. At the end of each month, we charge $0.46 per concept. You can delete unused models at any time for free.
Finally, using a custom model to predict works the same way described in the first paragraph above, except that you specify one of your own models to analyze the inputs. Instead of the $1.20 per thousand inputs it costs to do this with a public model, Custom Predicts cost $3.20 per thousand inputs.
Example 1: Predict (public model)
Let’s assume your application generates 1 million new images each month, and you want to analyze them using a public model. Your bill would appear as follows:
Example 2: Search
Let’s assume you have an image catalogue of 10 million images that grows by 1 million images each month. Let’s also assume that you perform 3 million searches over that catalogue each month. Your bill would appear as follows:
Example 3: Predict (multiple models)
Let’s assume your application generates 1 million new images each month. Let’s also assume that you want to analyze half of them with one of Clarifai’s public models, and half of them with your own 50-concept custom model that you trained using 75,000 inputs. Your bill would appear as follows:
Whew, that's it! We know that it's a departure from our previous pricing, but we're excited that it's a huge step forward in a positive direction for our developers. Please let us know if you have any pricing questions by shooting us an email at firstname.lastname@example.org.