An update to GPU-hour rates, effective Nov 9

No dark-mode styles — this email renders the same in a dark client.

New GPU-hour rates for custom model deployments and GPU rentals start November 9. Serverless API pricing stays the same.  ‌ ‌ ‌ ‌ ‌ ‌ ‌

https://deepinfra.com Pricing update

Thank you for building on DeepInfra. Many of you run production workloads on our dedicated GPUs, so we want to be upfront about a change: starting November 9, 2026, GPU-hour pricing is going up for custom model deployments and private GPU rentals.

Industry-wide increases in GPU and hardware costs have made new capacity more expensive. We’d rather adjust pricing than cut corners on capacity or reliability.

The new rates apply to all custom model deployments billed by uptime, including LLMs, embeddings, rerankers, audio and other model types, and to all private GPU rentals. This includes existing deployments and rentals, which move to the new rates automatically.

What isn’t changing: our rates remain among the most competitive in the market, and token-based pricing for serverless model APIs stays exactly where it is.
Private GPU pricing


** New GPU-hour rates effective November 9.
------------------------------------------------------------

Here are the new rates, per GPU-hour:
GPU  Memory Current From Nov 9
A100 80GB   $0.89   $1.49
H100 80GB   $2.20   $2.90
H200 141GB  $2.69   $3.49
B200 180GB  $3.69   $4.89
B300 270GB  $4.89   $5.89

Usage through November 8 is billed at current rates. Billing still works the same way, with per-minute granularity.
What it means for you


** No action needed.
------------------------------------------------------------

Custom model deployments: your deployments, autoscaling settings and endpoints stay the same and move to the new rates automatically on November 9. If you’d like to review your setup before then, you can adjust GPU type, count or scaling limits at any time.

GPU rentals: rentals already running, and any you start later, move to the new rates on November 9. Nothing else about your rentals changes.

If you have questions or want to talk through options for your workload, our team is happy to help.

See full pricing → (https://deepinfra.com/pricing)

Custom deployment docs → (https://docs.deepinfra.com/private-models/custom-llms)

Thank you for building with us.

— The DeepInfra team

Models (https://deepinfra.com/models)   ·   Docs (https://docs.deepinfra.com)   ·   Pricing (https://deepinfra.com/pricing)   ·   https://www.linkedin.com/company/deep-infra  ·   https://x.com/DeepInfra

You are receiving this because you have an account at deepinfra.com.
Deep Infra Inc . 2625 Middlefield Rd PMB 460 . Palo Alto, CA 94306-2516 . USA

Unsubscribe (https://deepinfra.us14.list-manage.com/unsubscribe?u=dedd07968badba1c4f3d5dc52&id=3c35f23c9d&t=b&e=5970bbdfef&c=07c2a7b0da)   ·   Update preferences (https://deepinfra.us14.list-manage.com/profile?u=dedd07968badba1c4f3d5dc52&id=3c35f23c9d&e=5970bbdfef&c=07c2a7b0da)

Accession

nse/deepinfra/2026-10-21/4716

From

DeepInfra · newsletter@deepinfra.com

Sent

21 October 2026 · 19:00 UTC