You’re viewing a demo workspace. Every number on this page is sample data. Sign in to use your own data

Inference FinOps

What each endpoint costs per token, how much of it sits idle, and where the money goes. Read-only from your provider billing and metrics.

Sep 24 – Oct 23, 2026

GPU spend, 30 days

$46k

5 endpoints

Tokens served

127.20B

Input and output, 30 days

Average utilization

56%

Spend-weighted

support-agent per 1M tokens

$0.392

Last 7 days, 6% lower than 4 weeks ago

Endpoints

Cost per 1M tokens uses GPU-hours billed, not GPU-hours busy.

Endpoint	Hardware	Utilization	Tokens/day	Per 1M tokens	Monthly	Idle/mo
support-agent Llama 3.1 70B	8× H100 CoreWeave · $2.79/GPU-hr	41%	1.48B	$0.367	$16,294	$9,613
doc-summarize Llama 3.1 70B	4× H100 AWS · $4.42/GPU-hr	83%	1.12B	$0.384	$12,906	$2,194
sql-copilot Qwen2.5 Coder 32B	4× H200 Lambda · $3.29/GPU-hr	68%	0.91B	$0.352	$9,607	$3,074
transcribe Whisper large-v3	2× H100 AWS · $4.42/GPU-hr	22%	0.42B	$0.512	$6,453	$5,033
moderation Llama Guard 3 8B	1× A100 80GB RunPod · $0.99/GPU-hr	57%	0.31B	$0.078	$723	$311

support-agentsql-copilotdoc-summarize

View as table

Day	support-agent	sql-copilot	doc-summarize
Sep 24	$0.413	$0.383	$0.431
Sep 25	$0.393	$0.385	$0.435
Sep 26	$0.472	$0.460	$0.503
Sep 27	$0.458	$0.437	$0.479
Sep 28	$0.402	$0.369	$0.404
Sep 29	$0.393	$0.377	$0.413
Sep 30	$0.391	$0.384	$0.419
Oct 1	$0.382	$0.376	$0.404
Oct 2	$0.382	$0.376	$0.407
Oct 3	$0.457	$0.444	$0.471
Oct 4	$0.445	$0.437	$0.474
Oct 5	$0.374	$0.369	$0.402
Oct 6	$0.387	$0.357	$0.402
Oct 7	$0.380	$0.362	$0.396
Oct 8	$0.386	$0.365	$0.403
Oct 9	$0.387	$0.370	$0.388
Oct 10	$0.441	$0.425	$0.479
Oct 11	$0.450	$0.425	$0.469
Oct 12	$0.371	$0.360	$0.399
Oct 13	$0.379	$0.361	$0.399
Oct 14	$0.375	$0.365	$0.382
Oct 15	$0.381	$0.366	$0.389
Oct 16	$0.374	$0.362	$0.383
Oct 17	$0.427	$0.419	$0.466
Oct 18	$0.449	$0.426	$0.468
Oct 19	$0.368	$0.361	$0.392
Oct 20	$0.381	$0.356	$0.400
Oct 21	$0.377	$0.366	$0.400
Oct 22	$0.369	$0.364	$0.399
Oct 23	$0.375	$0.360	$0.390