inferenstack — the market read on AI inference · last observation 2026-09-20. See the boards
About

Roadmap

An honest public roadmap — what's live, what's next, and what's deliberately an empty slot until we have the data. Gaps are stated with the reason, not hidden behind dimmed nav.

Inference — token prices

shipped
Price Spread
Same model across every provider — cheapest, median, spread, cache, tiers.
shipped
Blended Cost calculator
What a provider really costs your input:output + cache mix.
shipped
Value Quadrant
$/M blended vs coding score (Aider, seeded), fully interactive and free.
needs data
Service & media prices
Per-image / per-video-second / per-character. The page exists; the media-price ingest hasn't populated it yet, so it's disabled until it has data.
shipped
Free-Tier Board (interim)
Providers serving models for $0. Rate limits/caps need a free-tier ingest — that's still next.
needs data
Provider speed
Measured tok/s and TTFT per provider. Needs latency probes we don't run yet.

Compute — GPU rental

Models

Plan — decision engines

Data & API

shipped
7-day price history (free)
Public teaser window on every board.
next
30-day history (registered, free)
Sign in to see a month of history free.
next
Full history, CSV export, alerts
The paid tier: time-depth, export, automation, personalization.
next
Public API
Normalized prices and spreads as JSON — the sellable panel.

Hardware (exploratory)

later
GPU specs, hardware pages, benchmark corpus, used prices, build archetypes
Hardware is an input to fit today, not a destination. Exploratory tier.
How we label numbers →Methodology →