Dedicated GPUs by the hour and private LLM endpoints. Your models, your hardware, your data. Nothing shared, nothing logged.
Rent a GPU · Deploy an LLM · Docs · Pricing
Dedicated hardware by the hour, SSH access, full root. Bring any container, notebook, or CUDA stack.
| GPU | VRAM | From |
|---|---|---|
| H100 SXM | 80 GB | $2.49/hr |
| A100 | 80 GB | $1.29/hr |
| L40S | 48 GB | $0.90/hr |
B200 and H200 class also available. See full GPU pricing.
Pick a model, pick a GPU, get a private HTTPS endpoint. We handle driver, CUDA, and inference-engine alignment so your runtime starts correctly the first time.
| Model | Provider | Good for |
|---|---|---|
| Llama 3.3 70B | Meta | Complex reasoning, RAG, enterprise chat |
| Gemma 4 31B | General assistant workloads | |
| Nemotron-3 Nano | NVIDIA | Fast, low-cost reasoning and code |
| Devstral Small 24B | Mistral | Coding agents |
Change base_url and your key. Nothing else.
curl https://api.hexgrid.cloud/v1/chat/completions \
-H "Authorization: Bearer $HEXGRID_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "llama-3.3-70b-instruct",
"messages": [{"role": "user", "content": "Explain private GPU inference."}]
}'Pre-provisioned MiniMax H3, Wan 2.2, and LTX Video on a box only you can reach. Weights, custom nodes, and persistent output folders ready at launch.
- open-llm-benchmarks — TTFT, TPOT, throughput, latency, and cost-per-token for open models across dedicated GPUs. Reproducible, open methodology.
Partner datacenters are SOC 2 Type II, ISO 27001, and GDPR compliant. Regions in US, EU, and APAC. 200+ GPUs across providers.
hexgrid.cloud · Docs · Blog · Status · @hexgrid_cloud · HuggingFace · [email protected]
