Together AI H100 pricing
Together AI lists H100 from $3.99/GPU-hour for on-demand capacity. Check the minimum GPU count: the node total is the amount billed each hour.
Pricing at a glance
All Together AI H100 offers
More filters: provider, memory, node size, stock, sort, price unit
| GPU / provider | Billing | $/GPU-hour | Node · total per hour | Rent |
|---|---|---|---|---|
| H100 · Together AI80 GB | on-demand | $3.99 | – | Rent → |
| H100 · Together AI80 GB | spot / interruptible | $1.99 | – | Rent → |
Try another provider or reduce the minimum memory.
Tables rank by price, lowest first. Commercial relationships never affect ordering. How we collect and rank prices.
Same GPU, provider, billing and price on several hosts or listings: one row, with the count. – means the provider does not state it.
Monthly view assumes 730 hours of continuous use. Daily view assumes 24 hours.
The market median on-demand rate for H100 is $3.44/GPU-hour, counting each provider once. Together AI's lowest listed on-demand rate is 15.9% above it. Configurations and minimum node sizes differ.
What the Together AI bill includes
| Billing increment | per-hour |
|---|---|
| Minimum deposit | not published |
| Minimum purchase | 7 days for reserved GPU cluster tier |
| Data egress | not published |
| Storage | not published |
| When stopped | not published |
Billing terms updated 2026-07-07; confirm terms before renting. The GPU rate does not automatically include storage or data transfer.
Other H100 providers
More filters: provider, memory, node size, stock, sort, price unit
| GPU / provider | Billing | $/GPU-hour | Node · total per hour | Rent |
|---|---|---|---|---|
| H100 · Voltage Park80 GB | on-demand | $1.99 | – | Rent → |
| H100 SXM · TensorDock80 GB | on-demand | $2.25 | – | Rent → |
| H100 · Prime Intellect80 GB | on-demand | $2.43 | – | Rent → |
| H100 · Hyperstack80 GB | on-demand | $2.50 | – | Rent →2026-10-12Details
|
| H100 · GMI Cloud80 GB | on-demand | $2.60+30.0% in 30 days | – | Rent → |
Try another provider or reduce the minimum memory.
Tables rank by price, lowest first. Commercial relationships never affect ordering. How we collect and rank prices.
Same GPU, provider, billing and price on several hosts or listings: one row, with the count. – means the provider does not state it.
Monthly view assumes 730 hours of continuous use. Daily view assumes 24 hours.
All H100 rental rates · All Together AI GPUs · Estimate the total job cost