Skip to content
Hayonix
AI infrastructure, on your terms

Big ideas.
Serious compute.

Run open models, launch GPU workloads, and fine-tune on your data. One platform to take your next idea further.

Pay for what you use. No long-term commitment.

Conceptual illustration of graphite processor layers surrounding an emerald compute core.
  • OpenAI-compatible API
  • Per-second GPU billing
  • One prepaid balance
  • Built-in spend controls

One platform. More possibilities.

Start with an API.
Build on your terms.

From a single model call to a custom training run. Choose the right starting point, and grow from there.

/ 01

A model for your next idea

Inference API

Go from a prompt to a product. Build with open models through an API that fits the tools you already use.

  • Chat, embeddings, rerank and speech
  • Streaming and structured output
  • Shared models or dedicated endpoints

Usage-based pricing on shared models

Explore the model library
/ 02

Your workload. Your environment.

GPU compute

Bring your container and make yourself at home. Run a persistent pod or let serverless workers scale with demand.

  • GPU and CPU hardware options
  • SSH, web terminal and persistent volumes
  • Serverless workers that scale from zero

GPU pods billed by the second

Find your compute
/ 03

Make a model your own

Fine-tuning

Turn your data into a model that fits your task. Follow every run, keep your checkpoints, and deploy the result.

  • LoRA and full fine-tuning
  • Dataset validation and upfront estimates
  • Live metrics and downloadable checkpoints

Pay per training token

Explore fine-tuning

Made for developers

Familiar tools.
New possibilities.

Keep the OpenAI SDK you know. Point it at your inference endpoint, choose a model, and start building.

  1. 01

    Create your API key

    Keep access scoped to your project.

  2. 02

    Choose an open model

    Find the right fit in the model library.

  3. 03

    Make it part of your product

    Stream responses with the tools you use today.

Follow the quickstart
Your first requestOpenAI SDK
import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.hayonix.ai/v1",
    api_key=os.environ["API_KEY"],
)

response = client.chat.completions.create(
    model="MODEL_ID_FROM_THE_LIBRARY",
    messages=[
        {
            "role": "user",
            "content": "Summarise this incident report in three bullet points.",
        },
    ],
    max_tokens=512,
    temperature=0.7,
    stream=True,
    stream_options={"include_usage": True},
)
for chunk in response:
    if chunk.choices and chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)
print()
Add your API key and choose a model from the library.

Power without the guesswork

Serious hardware.
Straightforward pricing.

Pick your GPU. Run your workload. Pay for the seconds you use, with spending limits and usage you can track.

Explore all pricing
GPU computePay as you go

GPU prices are temporarily unavailable.

Check the pricing page

Prices in USD, excluding tax. Storage is billed separately.

A few things to know

Good questions.
Clear answers.

Find your starting point.

Explore the documentation
Should I start with inference or a GPU pod?

Use the inference API when you want to call a model without managing its infrastructure. Choose a GPU pod when you need your own container, dependencies, or development environment. Serverless workers are a good fit for jobs that arrive on demand.

Explore the guides
Can I keep using the OpenAI SDK?

Yes. Set the base URL to the inference API, add your project API key, and select a model from the library. Chat completions support streaming, tool calls, and structured output on compatible models.

Make your first request
How does billing work?

Add prepaid credit and use it across products. Text inference is charged by token, GPU pods by the second, and fine-tuning by training token. Speech, rerank, images, storage, and dedicated endpoints have their own meters. Review usage in the console and set spending limits and alerts.

See the full pricing breakdown
Can I bring my own models and containers?

Yes. Run your container image on a pod or serverless worker, bring a supported model from Hugging Face to a dedicated endpoint, or fine-tune a supported base model on your own dataset.

Launch your first pod

Your next chapter starts here

Less setup.
More building.

Bring your idea. We’ll bring the compute.