GPU · 312 TFLOPS dense BF16

A100 80GB fine-tuning cost estimates

All sixteen scenarios on the A100 80GB: GPU-hours, wall-clock on 8 GPUs and the cost range across 12 providers, at 35% MFU. Prices read 12 November 2026. Estimates, not quotes.

The figure

Where 312 TFLOPS comes from

“BFLOAT16 Tensor Core · 312 TFLOPS · 624 TFLOPS* ·”

nvidia.com

Scenarios

Cost by scenario

Scenario	GPU-hours	8 GPUs	Lowest	Median	Highest
7B LoRA fine-tune on 100 million tokens	7.12	53 min	$7.76	$13	$24
7B LoRA fine-tune on 1 billion tokens	71	8.9 h	$78	$135	$244
7B full fine-tune on 100 million tokens	11	1.3 h	$12	$20	$37
7B full fine-tune on 1 billion tokens	107	13 h	$116	$202	$367
13B LoRA fine-tune on 100 million tokens	13	1.7 h	$14	$25	$45
13B LoRA fine-tune on 1 billion tokens	132	17 h	$144	$251	$454
13B full fine-tune on 100 million tokens	20	2.5 h	$22	$38	$68
13B full fine-tune on 1 billion tokens	198	25 h	$216	$376	$681
34B LoRA fine-tune on 100 million tokens	35	4.3 h	$38	$66	$119
34B LoRA fine-tune on 1 billion tokens	346	43 h	$377	$656	$1,187
34B full fine-tune on 100 million tokens	52	6.5 h	$57	$98	$178
34B full fine-tune on 1 billion tokens	519	2.7 days	$566	$983	$1,780
70B LoRA fine-tune on 100 million tokens	71	8.9 h	$78	$135	$244
70B LoRA fine-tune on 1 billion tokens	712	3.7 days	$776	$1,350	$2,444
70B full fine-tune on 100 million tokens	107	13 h	$116	$202	$367
70B full fine-tune on 1 billion tokens	1,068	5.6 days	$1,165	$2,025	$3,665

Prices used

A100 80GB price by provider

Provider	$/GPU-hr	Configuration	Source
Thunder Compute	$1.09	1×	price page
Hyperstack	$1.35	1×	price page
Massed Compute	$1.35	1×	price page
Runcrateours, live	$1.73	1×	price page
Civo	$1.79	1×	price page
RunPod	$1.79	1× PCIe	price page
Crusoe	$2.00	1× PCIe	price page
Verda	$2.34	1× SXM	price page
Vultr	$2.40	1× PCIe	price page
CoreWeave	$2.70	8×	price page
OVHcloud	$3.07	1×	price page
Amazon Web Services	$3.43	8×	price page

Each provider's lowest listed on-demand configuration for this GPU. Marketplace figures are left out.

Questions

A100 80GB: common questions

What is the lowest cost scenario on the A100 80GB?

7B LoRA fine-tune on 100 million tokens: about $7.76 on Thunder Compute, 7.1 GPU-hours.

What throughput is assumed?

312 TFLOPS dense BF16 as printed by the maker, times 35% MFU, so about 109 TFLOPS of useful work.

How many providers list it?

12 with a current provider-set on-demand price. Runcrate is one of them at $1.73 live, marked as ours.

Disclosure

Fine-Tune Cost Lab is published by the Runcrate team, who also run Runcrate, an AI cloud. Runcrate is listed in the table for A100 80GB, marked "ours, live". It is sorted by price like every other row, with no pinned position. Other providers' prices are read from their own public pricing pages. We have no paid placement.