specimba

19.08 Update on your paid usage policy and your drained thousands of dollars while you are sleeping on that.
@akseljoonas @clem @lewtun
Have you ever checked your "https://endpoints.huggingface.co/ml-intern-explorers/audit-logs" ?
While some responsible person like me try to test models with afk stanby 15-30 minute auto turning off and not over GPU stacking for nonsense models, these two "https://huggingface.co/DedeProGames" and "https://huggingface.co/Sweaterdog" drained more than 6k dollar usage last month for at least seen.
"
Endpoints

$6,070.64

andy-4-2-wew
Nvidia L40S
9 hours and 54 minutes
$17.82
grape-2-pro-dvw
2x Nvidia RTX PRO 6000
33 minutes
$3.03
grape-2-pro-pgk
1x Nvidia H200
22 hours and 32 minutes
$112.67
grape-2-pro-pgk
2x Nvidia H200
309 hours and 15 minutes
$3,092.51
grape-2-pro-wrh
1x Nvidia H200
4 minutes
$0.33
grm-32-sky-v2
2x Nvidia RTX PRO 6000
311 hours and 44 minutes
$1,714.54
grm-32-sky-vllm
2x Nvidia RTX PRO 6000
41 minutes
$3.76
lrm-32-vllm
1x Nvidia RTX PRO 6000
70 hours and 9 minutes
$192.91
mira-1-xl-gcj
2x Nvidia A100
41 hours and 23 minutes
$206.92
mira-btq
Nvidia A100
47 hours and 20 minutes
$118.33
qwen-qwen3-5-27b-gguf-rlj
Nvidia L40S
1 hour and 14 minutes
$2.22
qwen3-6-40b-fable-fusion-6-c-vmw
4x Nvidia A100
1 hour and 13 minutes
$12.17
qwen3-8-27b-aeon-ultimate-un-hqg
Nvidia A100
11 hours and 28 minutes
$28.67
qwen3-8-27b-kce
2x Nvidia A100
78 hours and 7 minutes
$562.44
qwen3-8-27b-uncensored-fp8-nvb
Nvidia A100
56 minutes
$2.33
"
( yes that 28 from my model testing with qwen 3.8)
https://huggingface.co/organizations/ml-intern-explorers/settings/billing
can also be seen here
Current period usage
$11,724.99
Ends on
Sep 1

https://endpoints.huggingface.co/ml-intern-explorers/endpoints/grape-2-pro-pgk
Usage
Past Invoices
Instance Type Compute Time Rate (USD/hour) Cost (USD)
1x Nvidia H200
1352 minutes 5.00 112.67
2x Nvidia H200
18555 minutes 10.00 3,092.51
Total $ 3,205.17
Configuration
public
Anyone can access your endpoint, without authentication.
image-text-to-text
Created Aug 6 by DedeProGames • Last Edited Aug 19 at 8:35 PM by admin

https://endpoints.huggingface.co/ml-intern-explorers/endpoints/grm-32-sky-v2
Usage
Past Invoices
Instance Type Compute Time Rate (USD/hour) Cost (USD)
2x Nvidia RTX PRO 6000
18704 minutes 5.50 1,714.54
Total $ 1,714.54
Usage
Past Invoices
Instance Type Compute Time Rate (USD/hour) Cost (USD)
2x Nvidia RTX PRO 6000
18704 minutes 5.50 1,714.54
Total $ 1,714.54
Configuration
public
Anyone can access your endpoint, without authentication.
image-text-to-text
Created Aug 6 by DedeProGames • Last Edited Aug 19 at 2:58 AM by Sweaterdog

https://endpoints.huggingface.co/ml-intern-explorers/endpoints/qwen3-8-27b-kce
Usage
Past Invoices
Instance Type Compute Time Rate (USD/hour) Cost (USD)
2x Nvidia A100
4687 minutes 7.20 562.44
Total $ 562.44
Configuration
public
Anyone can access your endpoint, without authentication.
image-text-to-text
Created Aug 14 by DedeProGames • Last Edited Aug 18 at 10:00 PM by specimba

So, yes, I always try to involve them in drainers that have been paused for 30 minutes and are not being used. I went with a darker edit.
But, as I see now, after thousands of dollars in usage, when I was only involved in and used the 28-dollar latest Frontier model testing, the admin somehow got involved and miraculously zeroed out the balance after days and days of nonsense.

FYI...