Home/Large language models/Together AI

Together AI pricing

Together AI runs open-source AI models through APIs and offers GPU clusters, fine-tuning and dedicated hosting.

Large language modelsPriced per minuteSelf-serve + sales
Visit Together AIPrices checked Oct 29, 2026

How much does Together AI cost?

Usage price from $3/minute
Free plan
Not stated
Usage price
$3/minute
Enterprise
Contact sales
Read from www.together.ai/pricing on Oct 29, 2026. Price changed? Tell usSee the line each price came from

Watch out for

  • No refunds on a long commitment. Fees paid are non-refundable

How you buy and get started

Self-serve + salesButtons on their site: Start building, Contact Sales, Sign in, Get Started now
Sign up
Self-serve signup via API platform
"[Start building](https://api.together.ai/)"
Setup
Not stated
When sales steps in
Reserved capacity and some GPU hardware require contacting sales
"[Contact sales](/monthly-reserved)"

What Together AI does

Together AI is a cloud platform for running open-source models via serverless, batch, provisioned and dedicated inference, plus GPU clusters, sandboxes, storage, fine-tuning and post-training. Serverless usage is priced per million tokens and other products per GPU hour or per job.

Good for
Developers and teams that want to run, fine-tune or host open-source models at scale, from experimentation to production.
Channels
API, Playground, Together Chat, CLI
Main product
Serverless Inference
Also includes
Serverless Inference, Batch Inference, Provisioned Throughput, Dedicated Model Inference, Dedicated Container Inference, Together Link
Public API
Yes
MCP server
Not stated
  • Public API accessHigh-performance inference as APIs
  • Long context windowNot stated
  • Fine-tuning supportFine-tune open-source models for production workloads, using the latest research techniques.
  • Tool and function callingNot stated
  • Image and audio input[Chat](#)[Vision](#)[Image](#)[Audio](#)[Video](#)[Transcribe](#)[Embeddings](#)[Rerank](#)[Moderation](#)
  • SOC 2 complianceNot stated
  • Zero data retention optionBy choosing “No”, you are enabling Zero Data Retention (“ZDR”) for your use of the Services.
  • VPC or on-prem deploymentDeploy models on dedicated infrastructure.

The same 8 checks appear on every tool in Large language models. Each tick quotes the vendor's own site.

Security and your data

QuestionTheir answerWhere it says so
Trains AI on your data?no"We do not use any data collected from you to train our models without your explicit opt-in and consent."privacy
HIPAA (signs a BAA)Not stated
SOC 2Not stated
GDPRNot stated
CCPANot stated
PCINot stated
ISO 27001Not stated
Single sign-onNot stated
Trust centerNot stated

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

Every plan

PlanMonthlyBilled yearlyIncludedOver the allowanceWhere it says so
Kimi K3 (serverless)per 1M input tokens $3per minute – – – "| [Kimi K3](/models/kimi-k3) | $3.00 $0.30 (cached) | $15.00 |"
GLM-5.3 (serverless)per 1M input tokens $1.40per minute – – – "| [GLM-5.3](/models/glm-5-3) | $1.40 $0.26 (cached) | $4.40 |"
gpt-oss-120B (serverless)per 1M input tokens $0.15per minute – – – "| [gpt-oss-120B](/models/gpt-oss-120b) | $0.15 | $0.60 |"
Dedicated Inference NVIDIA HGX H100 on-demandper GPU per hour $5.49per minute – – – "| NVIDIA HGX H100 | $5.49 | [Contact sales](/monthly-reserved) |"
Dedicated Inference NVIDIA HGX B200 on-demandper GPU per hour $8.99per minute – – – "| NVIDIA HGX B200 | $8.99 | [Contact sales](/monthly-reserved) |"
GPU Clusters NVIDIA HGX H100 on-demandper GPU per hour $3.99per minute – – – "| NVIDIA HGX H100 | $3.99 |"
GPU Clusters NVIDIA HGX B200 on-demandper GPU per hour $8.19per minute – – – "| NVIDIA HGX B200 | $8.19 |"
Code Interpreter sessionper 60-minute session $0.03per minute – – – "| Session (60 minutes) | $0.03 |"
Shared Filesystem storageper GiB per month $0.16per minute – – – "| Shared Filesystem | $0.16 | GiB/month |"
Enterprise / reserved capacity Quote – – – "[Contact sales](/monthly-reserved)"

Fine print

Billing increment
Post-training billed per minute for each active model resource
"On-demand training infrastructure, billed per minute for each active model resource."
Billed beyond usage
Not stated
Contract
Term continues until use stops or Company terminates; you may not cancel without written consent
"You may not cancel or terminate this Agreement without our express written consent."
Refunds
Fees paid are non-refundable
"payment obligations are non-cancelable and non-pro-ratable for partial months, and fees paid are non-refundable."
Auto renew
Prepaid balance is auto-replenished when it falls below a threshold
"You authorize the Company to replenish your account by charging a minimum or the higher average monthly usage to Customer’s payment instrument each time Customer’s account balance "
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
Together Computer, Inc.
Address
251 Rhode Island Street, Suite 200, San Francisco, CA 94103
Online since
Dec 16, 2017
Domain registration date

Together AI alternatives

Reviews elsewhere

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

Questions

Is Together AI free?

No free plan is stated on Together AI's pricing page.

Does Together AI train AI on your data?

No, according to Together AI: "We do not use any data collected from you to train our models without your explicit opt-in and consent."