GPU · 181.05 TFLOPS dense BF16
L40 fine-tuning cost estimates
All sixteen scenarios on the L40: GPU-hours, wall-clock on 8 GPUs and the cost range across 6 providers, at 35% MFU. Prices read 12 November 2026. Estimates, not quotes.
The figure
Where 181.05 TFLOPS comes from
“BFLOAT16 Tensor Core TFLOPS 181.05 · 362.1**”
Scenarios
Cost by scenario
| Scenario | GPU-hours | 8 GPUs | Lowest | Median | Highest |
|---|---|---|---|---|---|
| 7B LoRA fine-tune on 100 million tokens | 12 | 1.5 h | $9.70 | $11 | $15 |
| 7B LoRA fine-tune on 1 billion tokens | 123 | 15 h | $97 | $114 | $155 |
| 7B full fine-tune on 100 million tokens | 18 | 2.3 h | $15 | $17 | $23 |
| 7B full fine-tune on 1 billion tokens | 184 | 23 h | $145 | $171 | $232 |
| 13B LoRA fine-tune on 100 million tokens | 23 | 2.8 h | $18 | $21 | $29 |
| 13B LoRA fine-tune on 1 billion tokens | 228 | 28 h | $180 | $212 | $287 |
| 13B full fine-tune on 100 million tokens | 34 | 4.3 h | $27 | $32 | $43 |
| 13B full fine-tune on 1 billion tokens | 342 | 43 h | $270 | $318 | $431 |
| 34B LoRA fine-tune on 100 million tokens | 60 | 7.5 h | $47 | $55 | $75 |
| 34B LoRA fine-tune on 1 billion tokens | 596 | 3.1 days | $471 | $554 | $751 |
| 34B full fine-tune on 100 million tokens | 89 | 11 h | $71 | $83 | $113 |
| 34B full fine-tune on 1 billion tokens | 894 | 4.7 days | $706 | $832 | $1,127 |
| 70B LoRA fine-tune on 100 million tokens | 123 | 15 h | $97 | $114 | $155 |
| 70B LoRA fine-tune on 1 billion tokens | 1,227 | 6.4 days | $970 | $1,141 | $1,547 |
| 70B full fine-tune on 100 million tokens | 184 | 23 h | $145 | $171 | $232 |
| 70B full fine-tune on 1 billion tokens | 1,841 | 9.6 days | $1,454 | $1,712 | $2,320 |
Prices used
L40 price by provider
| Provider | $/GPU-hr | Configuration | Source |
|---|---|---|---|
| Thunder Compute | $0.79 | 1× | price page |
| RunPod | $0.82 | 1× | price page |
| Massed Compute | $0.86 | 1× | price page |
| Hyperstack | $1.00 | 1× | price page |
| CoreWeave | $1.25 | 8× | price page |
| Runcrateours, live | $1.26 | 1× | price page |
Each provider's lowest listed on-demand configuration for this GPU. Marketplace figures are left out.
Questions
L40: common questions
What is the lowest cost scenario on the L40?
7B LoRA fine-tune on 100 million tokens: about $9.70 on Thunder Compute, 12 GPU-hours.
What throughput is assumed?
181.05 TFLOPS dense BF16 as printed by the maker, times 35% MFU, so about 63 TFLOPS of useful work.
How many providers list it?
6 with a current provider-set on-demand price. Runcrate is one of them at $1.26 live, marked as ours.
Related
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 L40, 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.