Ark Programmers
Standard
NVIDIA RTX 3090 (24GB VRAM)
20% OFF launch promo$160 $200 / month
- Ampere · GDDR6X
- Up to ~20B parameter models (Q4)
- Hugging Face .gguf models
- Private API endpoint
AI model hosting
Dedicated GPU hosting for your AI models
Pick one of seven NVIDIA GPUs, send us your Hugging Face .gguf model and get a private API endpoint. Every plan is 20% off during our launch. Not sure which GPU you need? Use the free sizing tool below.
New launch · AI model hosting
Host your own AI model on dedicated GPUs
Ark Programmers now hosts Hugging Face .gguf models on dedicated GPUs, billed monthly. Pick a GPU, send us your model, and get a private API endpoint. The H100 runs models up to 200B parameters.
Fast
NVIDIA RTX 4090 (24GB VRAM)
20% OFF launch promo$200 $250 / month
- Ada Lovelace · GDDR6X
- Up to ~20B models, faster tokens/sec
- Hugging Face .gguf models
- Private API endpoint
Fastest (consumer)
NVIDIA RTX 5090 (32GB VRAM)
20% OFF launch promo$240 $300 / month
- Blackwell · GDDR7
- Up to ~32B parameter models
- Hugging Face .gguf models
- Private API endpoint
Big VRAM
NVIDIA RTX A6000 (48GB VRAM)
20% OFF launch promo$280 $350 / month
- Ampere · ECC GDDR6
- Up to ~70B models (Q4)
- Hugging Face .gguf models
- Private API endpoint
Datacenter
NVIDIA A100 (80GB VRAM)
20% OFF launch promo$320 $400 / month
- Ampere · HBM2e
- 70B+ models, high throughput
- Hugging Face .gguf models
- Private API endpoint
Workstation (Blackwell)
NVIDIA RTX 6000 PRO (96GB VRAM)
20% OFF launch promo$360 $450 / month
- Blackwell · GDDR7 ECC
- Up to ~120B+ models
- Hugging Face .gguf models
- Private API endpoint
Datacenter (max)
NVIDIA H100 (80GB VRAM)
20% OFF launch promo$400 $500 / month
- Hopper · HBM3
- Up to 200B parameter models
- Hugging Face .gguf models
- Private API endpoint
Which GPU does my model need?
Enter a Hugging Face model and our AI recommends a GPU and explains its memory needs.
Need a business logo, ads for your business or a consultation?
We design logos, create ads for your business and offer one-to-one consultations. Contact us through our CRMP.