Provider guide · Model APIs / GPU machines

Together AI

Managed model APIs, dedicated inference and GPU clusters.

Source checked 11 Oct 2026 · partial product record

We currently earn no commission from these links. How we fund this site

View Together AI pricing Explore 2 price records ↓

Good for: Explore model APIs or GPU infrastructure, depending on the product.
You manage: APIs need application integration; GPU machines need workload operations.
GPUs listed: H100 SXM · H200 SXM · B200 SXM 180GB
Price basis: 2 current compute rates · checked 14 Oct 2026
Not confirmed here: Live GPU availability, capacity in your region and your complete bill (storage, transfer, tax).

Understand the price

2 recorded configurations. Pick the one that fits.

Compare operating models →

Recorded compute prices help you estimate a workload. Availability and a complete bill still need checking with Together AI.

We currently earn no commission from these links. How we fund this site

8× H100 cluster entry configuration

8 GPUs · GPU cluster

NVIDIA H100 SXM 80GB · on-demand
Region not confirmed · Availability not checked

Price checked 14 Oct 2026 · valid until 21 Oct 2026, 00:39 UTC
GPU variant inferred from the provider’s official specifications

$31.92 / hour

USD · 8 GPUs · full configuration

Compute only · extras excluded

$3.99 per GPU-hour

1 hour$31.92

8-hour day$255.36

Compute only. Estimate: a billing rule is not published.

Estimate my monthly cost →Check this price at Together AI

Configuration notes & source

Included

GPU cluster compute

Costs to check

Applicable taxes; Extra storage and optional services

Source notes

Cluster product, not the separate dedicated inference price.

Product page lists H100 scale from 8 to 256 GPUs.

The pricing pages checked do not publish a CPU or RAM allocation for GPU clusters.

Together lists NVIDIA HGX H100; HGX boards use SXM modules.

Official pricing source ↗

256× H200 cluster entry configuration

256 GPUs · GPU cluster

NVIDIA H200 SXM 141GB · on-demand
Region not confirmed · Availability not checked

Price checked 14 Oct 2026 · valid until 21 Oct 2026, 00:39 UTC
GPU variant inferred from the provider’s official specifications

$1,533.44 / hour

USD · 256 GPUs · full configuration

Compute only · extras excluded

$5.99 per GPU-hour

1 hour$1,533.44

8-hour day$12,267.52

Compute only. Estimate: a billing rule is not published.

Estimate my monthly cost →Check this price at Together AI

Configuration notes & source

Included

GPU cluster compute

Costs to check

Applicable taxes; Extra storage and optional services

Source notes

Product page lists H200 scale from 256 to 1,000 GPUs.

A per-GPU rate must never be presented as a single-GPU rental offer.

The pricing pages checked do not publish a CPU or RAM allocation for GPU clusters.

Together lists NVIDIA HGX H200; HGX boards use SXM modules.

Official pricing source ↗

Know when the meter is running.

Cluster rates and dedicated-inference rates are separate products. H100 cluster entry configuration is 8 GPUs; H200 page lists 256 GPUs.

Selected model APIs support image/video generation and agent tool calling (reviewed 12 October 2026). The application must execute tool calls and manage its agent workflow; model inference alone is not application hosting.

Explore the setup

Check the product behind the name.

Products in this directory: Together AI model APIsModel APIsView product (opens in a new tab)
Together AI GPU infrastructureGPU machinesView product (opens in a new tab)
GPU hardware in this catalogue: H100 SXM · H200 SXM · B200 SXM 180GB
Region and availability: Confirm your selected product and region directly. Live capacity has not been checked.

Sources: official product information ↗ · official pricing ↗
Source review does not establish live availability or a commercial partnership.