Open Models, Deployed in Minutes
Run Llama, DeepSeek, Phi, and Gemma as auto-scaling inference endpoints with an OpenAI-compatible API — hosted on sovereign infrastructure in the region you choose.
Llama 3.3 70B
ENDPOINT • UK-SOUTH
curl https://api.gravitystack.cloud/v1/chat/completions \
-H "Authorization: Bearer $GS_API_KEY" \
-d '{
"model": "meta-llama/Llama-3.3-70b",
"messages": [{"role": "user",
"content": "Hello!"}]
}'Time to First Token
180 ms
Throughput
95 tok/s
Production Inference, Zero Ops
Everything you need to ship Gen AI features — without managing GPUs, drivers, or serving stacks.
OpenAI-Compatible API
Point your existing SDKs at our endpoint and switch models with one line. No rewrites, no vendor lock-in.
Auto-Scaling Endpoints
Serverless endpoints scale with your traffic — from zero to production load — and you only pay for the tokens you use.
Sovereign & Private
Prompts and completions stay in your chosen region. No training on your data, ever — and zero egress fees.
A Curated Model Catalog
The best open-weight models, benchmarked and ready to deploy.
Llama 3.3 70B
Latest Llama model with improved reasoning
DeepSeek-R1 70B
Open reasoning model for complex tasks
Phi-4 14B
Small model with big-model quality
Gemma 2 9B
Efficient language model for high-throughput use
Llama 3.2 3B
Lightweight Llama for fast, cheap inference
Ship Your First Gen AI Feature Today
Deploy a serverless endpoint in minutes and start building with the best open models.