Observed arrival · 2026-10-26
InferCPI puts AI inference costs on the record
A weekly index tracking AI model prices, estimated serving margins, cost per task, usage mix, and compute spending.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Researchers comparing AI model inference costs
- Worth noticing
- Serving-cost estimates bracket GPU rent from a fully loaded Vast.ai H100 to AWS on-demand at 40% utilization.
Field notes
Several views keep different questions separate: same-model price changes are chained apart from basket turnover, while the task-cost view uses eight short tasks with checkable answers and reports charges per 1,000 tasks. Open-weight models are compared at both lab and third-party-host prices. The page also marks important limits: desk-machine speeds are estimates, and chip-company revenue is used as an upper bound rather than a direct measure of AI compute sales.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue