Home/AI API platforms/DeepInfra

DeepInfra pricing

DeepInfra is an inference API that serves 100+ open and third-party models, billed per token or by execution time, plus rented GPUs.

AI API platformsPriced per minuteSelf-serve + sales
Visit DeepInfraPrices checked Oct 29, 2026

How much does DeepInfra cost?

Usage price from $0.06/minute
Free plan
Not stated
Usage price
$0.06/minute
Enterprise
Contact sales
Read from deepinfra.com/pricing on Oct 29, 2026. Price changed? Tell usSee the line each price came from

In your favor

  • No long-term contract. no long-term contracts
  • No setup fee. no upfront costs

How you buy and get started

Self-serve + salesButtons on their site: Let's Go, Book a consultation, Contact Sales, Log In
Sign up
Add a card or pre-pay to use the services
"You have to add a card or pre-pay or you won't be able to use our services."
Setup
Not stated
When sales steps in
Accounts limited to 200 concurrent requests; contact for more
"We limit each account to 200 concurrent requests. If you want more drop us a line"
Support
Hands-on technical support
"We are there for you with our simple APIs and hands-on technical support."

What DeepInfra does

DeepInfra runs text, image, speech, embedding and other models on its own US data centers behind pay-as-you-go APIs. It also offers custom LLM deployments on dedicated GPUs and dedicated clusters through sales.

Good for
Developers and companies that want low-cost API access to many models, from startups to enterprises, with zero data retention.
Channels
API, Web chat, Dashboard
Main product
AI inference API
Also includes
100+ models, Per-token pricing for language models, Cached input pricing on many models, Service tiers: Standard, Priority, Flex, Custom LLM deployments on dedicated GPUs, Embeddings, speech, image models
Public API
Yes
MCP server
Not stated
  • Pay-per-token pricingSome of our langauge models offer per token pricing.
  • Multiple models availableDeepInfra provides 100+ models to cover all your needs.
  • Fine-tuning supportNot stated
  • Streaming responsesNVIDIA Nemotron speech models for streaming multilingual transcription and speaker diarization.
  • Tool and function callingadds reasoning and tool use on top.
  • SOC 2 complianceDeepInfra is SOC 2 and ISO 27001 certified.
  • No training on your datawill not use Customer Data to train, fine-tune, or otherwise improve any model
  • Discounted batch processingNot stated

The same 8 checks appear on every tool in AI API platforms. Each tick quotes the vendor's own site.

Security and your data

QuestionTheir answerWhere it says so
Trains AI on your data?no"We will not store, sell, or train using this data unless we have your explicit consent."privacy
HIPAA (signs a BAA)Not stated
SOC 2Compliant"SOC 2 Compliant"security
GDPRIn progress"GDPR In progress"security
CCPANot stated
PCINot stated
ISO 27001Compliant"ISO 27001 Compliant"security
Single sign-onNot stated
Trust centerhttps://trust.deepinfra.com"[Trust Center](https://trust.deepinfra.com)"home

Security answers are the vendor's own claims, quoted. We don't audit them.

Every plan

PlanMonthlyBilled yearlyIncludedOver the allowanceWhere it says so
Token pricing (example: DeepSeek-V4-Flash-0731)per 1M input tokens $0.06per minute – – – "| [DeepSeek-V4-Flash-0731](/deepseek-ai/DeepSeek-V4-Flash-0731) | 1024k | $0.06 / $0.015 c"
Custom LLM GPU: A100 80GBper GPU-hour $1.49per minute – – – "| A100 | 80GB | **$1.49** / GPU-hour |"
Custom LLM GPU: H100 80GBper GPU-hour $2.90per minute – – – "| H100 | 80GB | **$2.90** / GPU-hour |"
Custom LLM GPU: H200 141GBper GPU-hour $3.49per minute – – – "| H200 | 141GB | **$3.49** / GPU-hour |"
Custom LLM GPU: B200 180GBper GPU-hour $4.89per minute – – – "| B200 | 180GB | **$4.89** / GPU-hour |"
Custom LLM GPU: B300 270GBper GPU-hour $5.89per minute – – – "| B300 | 270GB | **$5.89** / GPU-hour |"
DeepCluster B300 (5-year term)per GPU-hour $1.98per gpus (minimum) – 256 GPUs (minimum) – "NVIDIA B300 · 5-year term $1.98/GPU-hr"
Dedicated instances and clusters Quote – – – "For dedicated instances, DGX H100, B200 and B300 clusters with 3.2Tbps bandwidth, please c"

Fine print

Billing increment
Custom LLM GPUs billed in minute granularity
"Billed in minute granularity"
Billed beyond usage
Not stated
Contract
no long-term contracts
"Low pay-as-you-go pricing - no long-term contracts, no hidden fees, no surprises."
Refunds
Not stated
Auto renew
Not stated
Price change notice
Not stated

"Not stated" means we read their pricing page and terms and they don't say.

About the company

Legal name
Deep Infra Inc.
Address
2625 Middlefield Road #460
Online since
Dec 8, 2017
Domain registration date
Founder
Nikola Borisov, Founder and CEO
Founder
Georgios Papoutsis, Founder
Founder
Yessenzhar Kanapin, Founder
Funding
$107M Series B, May 4, 2026, co-led by 500 Global and Georges Harik
"We've raised $107M Series B to scale the inference cloud"
Customers
"25x token volume growth since Series A"
their claim
Backing
"Co-led by 500 Global and Georges Harik, with participation from A.Capital Ventur"
their words

DeepInfra alternatives

xAI API$8 /moFree plan · checked Oct 29, 2026Hugging Face Inference Providers$9 /moFree plan · checked Oct 29, 2026Anthropic Claude API$17 /moFree plan · checked Oct 29, 2026AI/ML API$20 /minutechecked Oct 29, 2026Azure OpenAI ServiceQuotechecked Oct 29, 2026Baseten$1.40 /minutechecked Oct 29, 2026Cerebras InferenceQuotechecked Oct 29, 2026Cloudflare Workers AI$0.01 /1,000 neuronsFree plan · checked Oct 29, 2026

Reviews elsewhere

G2 ↗Trustpilot ↗Reddit ↗Product Hunt ↗

We link to reviews instead of copying ratings: those sites don't allow reuse, and we don't test products ourselves.

Questions

Is DeepInfra free?

No free plan is stated on DeepInfra's pricing page.

Does DeepInfra train AI on your data?

No, according to DeepInfra: "We will not store, sell, or train using this data unless we have your explicit consent."