Power your AI, scale on your terms

Start instantly with full API access and our complete model suite. Validate in the cloud, then move on-prem when you're ready - fully supported, fully yours.

Full flex with every credit

Balance = credits. Simple as that. 1 USD = 1 million credits, spendable across all models and features. No plan lock-ins, no resource silos - just total control over how and when you use your compute.

Secure payments, no surprises

All payments go through Stripe - encrypted, bank-level security, no shady stuff. No auto-renewals, no hidden charges. Just pay once and use when you want.

Unlimited time to use your credits

Prepaid = predictable. No monthly bills, no surprise charges. Top up once and take however long you need to spend your credits on any model, any endpoint.

Jumpstart

5 USD = 5 000 000 credits

Kick the tires. Try ARK on your terms.

Perfect for:

exploring LLMs & generative AI without the heavy lift

Go to ARK Cloud

Command

500 USD = 500 000 000 credits

Big credits. Bigger control.

Perfect for:

production-grade inference, ready to go

Go to ARK Cloud

Explore model-level pricing & see how far your credits go

LLMs

All prices shown are per million tokens in USD

Stateful Stateless
Input token price Output token price Input token price Output token price Function calling JSON Mode
meta-llama/Llama-3.1-8B-Instruct 0.01 0.49 0.29 0.29
meta-llama/Llama-3.3-70B-Instruct 0.01 4.99 2.79 2.79
speakleash/Bielik-11B-v2.6-Instruct 0.01 0.99 0.49 0.49 -

Embedding model

Text embeddings for semantic search and similarity

Price
BAAI/bge-m3
$0.01 per million tokens

Image generation

AI-powered image creation and generation

Price
stabilityai/stable-diffusion-3.5-large
$0.019 per image

Automatic Speech Recognition

Convert speech to text with high accuracy

Price
openai/whisper-large-v3-turbo
$0.0144 per hour transcribed

Efficient inference of any model

Custom model hosting and inference solutions

Price
Any Open Source model On request
Your model On request

Questions about pricing or need something custom?

Talk to real engineers. We'll help you pick the right tier, map out a custom deployment, and build terms that match your roadmap.