To increase speed and reduce your costs, we are continually testing smaller AI models. We make these models available to you the moment they have reached the Alpha stage. This page lists the currently available models and their respective stage.
To specify a model, simply add {model: "name_of_model"} to the data portion of your POST request.
gpt-4. Currently in Beta.
It is charged at 100% of the token cost.
gpt-4o. Currently in Alpha.
It is charged at 5% of the token cost. This means that for every 100 tokens sent/received from this model, we only deduct 5 tokens from your balance. This model is also more than twice the speed of the gpt-4 model.
How Models are Tested
We test models using approximately 150 queries that are known to produce hallucinations in LLMs. If the model achieves 100% accuracy, it is then designated an alpha model, and is made available to you.
We then test the model on approximately 1,000 queries. Once the model achieves 100% accuracy, it is then designated a beta model.
We then test the model on approximately 2,500 queries. Once the model achieves 100% accuracy, it is then designated a production model.
Continually check back to this page to see the available models and their respective costs. We are working hard to make 100% accurate AI both fast and affordable.