DeepInfra Custom LLMs
Dedicated hosting on DeepInfra for a customer's own model: the model runs on dedicated NVIDIA A100, H100, H200, B200 or B300 GPUs with automatic scaling, billed by the minute and invoiced weekly at a per-GPU-hour rate (A100 $1.49 to B300 $5.89 on the page read).
Find alternatives to DeepInfra Custom LLMs
What it does
- Model hosting
- Model inference
- GPU cloud
Features 2
- Deploy your own modelLets the customer deploy their own model on DeepInfra hardware and pay for uptime
- Automatic scalingAutomatically scales a deployed model to more hardware as load requires
Sources
- Pricing
-
Pricing page, Custom LLMs: "pay for uptime", A100 80GB "$1.49 / GPU-hour", H100 "$2.90", H200 "$3.49", B200 "$4.89", B300 "$5.89"; "Billed in minute granularity", "Invoiced weekly".
- Description
-
Official documentation · 8 Nov 2026
Pricing page: "You can deploy your own model on our hardware and pay for uptime. You get dedicated SXM-connected GPUs (for multi-GPU setups), automatic scaling to handle load fluctuations and a very competitive price."
- What it does
-
Official documentation · 8 Nov 2026
Pricing page: "You can deploy your own model on our hardware and pay for uptime. You get dedicated SXM-connected GPUs (for multi-GPU setups), automatic scaling to handle load fluctuations and a very competitive price."
- Sold within
-
Pricing page publishes the per-GPU-hour rates for custom deployments.
- Name
-
Official documentation · 8 Nov 2026
Name as published on the product page.
- URL
-
Official documentation · 8 Nov 2026
Product page, read 2026-11-08.
- Deployment
-
Official documentation · 8 Nov 2026
Vendor-run dedicated hosting; no self-hosted option is named.
- Interfaces
-
Official documentation · 8 Nov 2026
Vendor-run dedicated hosting; no self-hosted option is named.
Also from DeepInfra
-
DeepInfra Inference API Model API
Hosted API serving open-weight models for text generation, embeddings, reranking, image, video, speech and music, billed per token for language models and by inference time for most others, with a batch API, private and custom-model deployment and LoRA adapters; per-token rates are listed on the pricing page, with no long-term contracts or upfront costs.
-
DeepInfra GPU Instances Infrastructure service
On-demand NVIDIA GPU containers from DeepInfra, started in under ten seconds with a chosen base image such as PyTorch or CUDA and reached over SSH, for training, fine-tuning or running inference, billed by the hour with no minimums (a single B200 at $4.89 an hour on the page read).
-
DeepCluster Infrastructure service
A dedicated NVIDIA Blackwell B300 GPU cluster that DeepInfra designs, procures, deploys in a Tier 3 datacenter and operates for the customer, who owns the hardware, on a three- to five-year contract; the page quotes $2.99 per GPU-hour all-in on a three-year term and 256 to 5,000 GPUs, through sales.
Open DeepInfra in the directory
Something wrong here? Send a correction — quote this product id: deepinfra-custom-llm-deployments.