FAQ
What is kluster.ai?
kluster.ai is a decentralized, high-performance compute network that provides developers and enterprises with tools to run large-scale batch inference jobs at a fraction of the cost compared to compared to traditional cloud providers.
What is the mission of kluster.ai?
Our mission is to empower AI creators to focus on what they do best—building transformative technologies—while we manage the heavy lifting of compute. kluster.ai was born out of the need to make large-scale AI compute accessible, efficient, and affordable. As we saw more companies and researchers pushing the limits of AI development, we noticed that the cost and complexity of compute were becoming bottlenecks to innovation. We are here to change that.
Why should I use kluster.ai?
kluster.ai is a cutting-edge AI compute infrastructure designed to provide large-scale, affordable GPUs for your workloads. Because it’s serverless, we remove the hassle of managing, maintaining, and scaling your own infrastructure.
How can I get started with kluster.ai?
To run batch inference workloads, send requests to our OpenAI API-compatible endpoint. Sign up at platform.kluster.ai or refer to our getting started guide.
What use cases can I use kluster.ai for?
Currently, we support batch inference use cases like summarization, classification, sentiment analysis, entity extraction, data formatting, text optimization, and much more.
Are you API-compatible with the OpenAI library?
Yes, kluster.ai is API-compatible with the OpenAI library. Currently, we support model, messages, and stream functions. If additional request properties are needed, they can be requested during the Early Access Plan.
How much does kluster.ai cost?
During the early access period we provide $500 credits in your account to help you get started. By the end of the Early Access Program, we’ll communicate our go-live pricing.
What’s the difference between the Llama models and the kluster.ai optimized models?
Currently, we support optimized versions of the Llama 3.1 8B, 70B, and 405B models that enhance inference speed without sacrificing accuracy at the lowest cost. We are planning to support more LLMs in future and would love to hear which models you would like us to support. Let us know by contacting [email protected].
Can I fine tune the models on kluster.ai?
Currently, we do not support fine-tuning of the models. However, it’s on our roadmap and we’d love to speak to you about your use case. Drop a message to [email protected] to arrange a follow-up call.
Will kluster.ai use the inputs or outputs to train models?
No. We won’t use any of your data to train models. Our mission is to make large-scale AI compute accessible, efficient, and affordable. Creating our own models falls outside of our core business interest.
What support is provided during the Early Access Program?
During the early access program, you can contact us at [email protected] describing your issue. We’ll respond as soon as possible, within 1 business day at the latest.
Why don’t you support the Llama reference models?
We don’t support the Llama reference models because our kluster.ai optimized models closely match the accuracy of the BF16 reference models but with much higher performance.
Where are your suppliers located?
Currently, our suppliers and based in USA and Canada.