From $0.01 usage-based · Free trial🇺🇸
US inference platform for open, fine-tuned and custom models on autoscaling GPUs, with Model APIs, training and self-hosting.
DeepInfra AI features and pricing (October 2026)
Inference API for more than 100 open and commercial models billed per token, plus GPU instances and dedicated clusters.
Usage-based
from $0.02 usage-based
Pricing details
Usage-based billing with no long-term contract, prepaid or invoiced monthly; token prices from $0.019 per million input tokens on the pricing page, dedicated H100 at $2.20 per GPU hour, as of September 2026.
What is DeepInfra?
DeepInfra offers an inference API for more than 100 AI models with per-token billing, plus GPU instances and dedicated clusters, for developers who want to run models without managing servers.
The catalog covers text generation, image synthesis, speech and embeddings, including open models and models from Google and Anthropic. GPU instances and the DeepCluster product add dedicated capacity. The company is based in Palo Alto, states that it runs its own US data centers, and raised a Series B round in May 2026. According to its documentation, inputs and outputs are held only in memory during standard inference and are not used for training.
DeepInfra AI features and use cases
- Calls many models through one APIDevelopers access open and commercial models with one account and pay per token. Switching models for a feature needs only a changed model name.
- Processes text without retentionInputs and outputs are not written to disk during standard inference, and only metadata such as request ID and cost is logged. This lowers the risk when prompts contain internal information.
- Runs custom models on dedicated GPUsDedicated GPUs and clusters host a company's own or fine-tuned models. Teams with steady traffic get fixed capacity at hourly rates instead of per-token pricing.
What sets DeepInfra apart
- Operates its own US data centers instead of reselling hyperscaler capacity, according to the vendor.
- Spending limits can be set per account to cap usage-based costs.
Where DeepInfra is strongest
United States Available internationally
Deep Infra Inc. is based in Palo Alto, California, according to its privacy policy, and operates its own data centers in the United States.
Questions to ask before you buy DeepInfra
Requests to Google or Anthropic models are forwarded to those vendors and follow their policies, so check which models you use for personal data. Processing takes place in the United States. A data processing agreement is not published, so contact the data protection officer. Bulk inference may keep data longer on encrypted disk.
DeepInfra reviews
No reviews yet. Be the first to share how your team uses it.
Every review shows its verification level. We read each one before it goes live. How we check reviews
Questions about DeepInfra
Does DeepInfra store my prompts?
Not in standard inference: inputs and outputs stay in memory and only metadata is logged. Exceptions apply to Google and Anthropic models and to bulk requests.
How is DeepInfra billed?
By usage with no long-term contract, either prepaid or invoiced monthly. Spending limits can be set in the account.
Is there a free tier?
The pricing page did not list a free tier in September 2026.
DeepInfra update historyno updates yet
No changes recorded for DeepInfra yet.
Changes the vendor made and confirmed user reports, each checked against the vendor's own pages. We review every tool monthly.
Alternatives to DeepInfra
Tools in the same category (Serverless inference) with the most overlapping markets.
From $250 per month · Free plan🇺🇸
Serverless Python platform for AI inference, training, batch jobs and sandboxes with per-second GPU billing.
Usage-based🇺🇸
GPU cloud with low-cost hourly Pods, serverless endpoints that scale to zero and clusters for developers and small teams.
From $1.50 usage-based🇺🇸
US AI cloud on its own data centers, with NVIDIA and AMD GPUs, managed Kubernetes and Slurm, and managed inference.