Insights

Company Updates & Technology Articles

Reinforcement learning from human feedback (RLHF)

Reinforcement learning from human feedback (RLHF)

Reinforcement learning from human feedback (RLHF) is a machine learning approach that leverages a combination of human feedback and reinforcement learning to train AI models.

From Data to Decisions: A Guide to the Core AI Technologies

AI is not one thing. It is a constellation of several technologies that enables machines to perceive, understand, act and learn with human-like intelligence.

How to build a generative AI solution

With generative AI, companies can unlock unprecedented levels of innovation, efficiency, speed, and accuracy, creating an unbeatable advantage in today’s hyper-competitive marketplace.

Demystifying AI: Practical Use Cases and Applications Across Key Industries

Unlock the potential of AI for your business. Learn about its practical applications across industries in this comprehensive article.

How to build a private LLM?

Language models are the backbone of natural language processing (NLP) and have changed how we interact with language and technology.

ZBrain credits and their usage

A credit is a unit of usage on ZBrain. Credits are consumed whenever you perform actions such as embedding documents or querying an app. Here’s how credits are utilized:

Document Embedding:

When you upload a document, it is processed and embedded to create a searchable vector representation. This step consumes credits based on the size of the document and the embedding model used.

Input Processing:

When you ask a question, the relevant context is retrieved from the embeddings and processed by the Large Language Model (LLM). This step incurs a credit cost based on the model and the number of input tokens.

Generating Output

The LLM then generates a response based on the processed input, consuming additional credits based on the number of output tokens.

The cost in credits for various models and processes is detailed below:

Models and Credit Consumption:

Model	Input Cost	Output Cost
GPT-4o	2,500 credits / 1M tokens	7,500 credits / 1M tokens
GPT-4	15,000 credits / 1M tokens	30,000 credits / 1M tokens
GPT-4-32k	30,000 credits / 1M tokens	60,000 credits / 1M tokens
GPT-3.5 Turbo	250 credits / 1M tokens	750 credits / 1M tokens

Embedding Models:

Model	Input Cost
text-embedding-3-small	20 credits / 1M tokens
text-embedding-3-large	130 credits / 1M tokens
ada v2	100 credits / 1M tokens

When you create a knowledge base or query an app, credits are deducted based on the embedding, input, and output token usage. This ensures you only pay for what you use while leveraging the full power of advanced AI models.