Models · Google · live prices
Gemma 2 27B: GPUs needed and cost to self-host
27.2B parameters. In FP8 that is about 33 GB with room for the KV cache.
Memory, FP8
33 GB
weights + ~20%
Cheapest per 1M output tokens
$0.09
1× A100 · on-demand · 60% utilization
Hosted API, sample list price
$0.80
per 1M output tokens
Cost per 1M output tokens, FP8
Prices observed 11m ago. Cheapest on-demand offer per GPU, whole instances, 60% average utilization.
- A10080 GB 1× on Vast.ai · on-demand · $0.34/hr · ~1,799 tok/s $0.09per 1M out
- H200141 GB 1× on Vast.ai · on-demand · $1.32/hr · ~4,235 tok/s $0.14per 1M out
- 509032 GB 2× on Vast.ai · on-demand · $0.92/hr · ~1,897 tok/s $0.22per 1M out
- 409024 GB 2× on Vast.ai · on-demand · $0.62/hr · ~1,067 tok/s $0.27per 1M out
- RTX PRO 600096 GB 1× on Vast.ai · on-demand · $1.00/hr · ~1,581 tok/s $0.29per 1M out
- H10080 GB 1× on RunPod · community · $1.99/hr · ~2,956 tok/s $0.31per 1M out
- B300288 GB 1× on RunPod · community · $6.94/hr · ~7,059 tok/s $0.46per 1M out
The tick marks the sample hosted-API price. Estimates: tokens/s from memory bandwidth with batched serving, not a benchmark. Idle hours still bill, so lower utilization means pricier tokens.
GPUs per replica by precision
One node, split 1, 2, 4 or 8 ways; 90% of VRAM usable.
| GPU | BF1665 GB | FP833 GB | INT416 GB |
|---|---|---|---|
| H10080 GB | 1× | 1× | 1× |
| H200141 GB | 1× | 1× | 1× |
| B300288 GB | 1× | 1× | 1× |
| A10080 GB | 1× | 1× | 1× |
| RTX PRO 600096 GB | 1× | 1× | 1× |
| 509032 GB | 4× | 2× | 1× |
| 409024 GB | 4× | 2× | 1× |