Find the cheapest way to run AI.

Compare LLM APIs, GPU clouds, vector databases and LLMOps tools - normalised and re-checked every day.

Then calculate what your workload actually costs.

Calculate & build

Stack cost

Price a whole AI stack at once - LLM, embeddings, vector DB, storage - and see what to change to cut it.

one number, layer by layer →

Six focused calculators: API cost, self-host break-even, fine-tuning, vector index, GPU ROI.

6 calculators →

Build your own AI infrastructure: a local workstation or a private GPU server.

6 reference builds →

Compare

Token prices for every major model API, normalised and updated daily.

from $0.06 / 1M tokens →

Hourly GPU rental across specialised clouds - on-demand and spot.

from $0.08 / GPU-hr →

Vector databases, embedding APIs and object storage, with egress fees flagged.

vector DB from $18 / mo →

The LLMOps landscape: observability, gateways, evals, guardrails.

29 tools · 18 open source →

What a GPU you own earns rented out, and how fast it pays for itself.

RTX 5090 · ~14 mo payback →

Learn

Plain definitions of every concept behind the numbers - and what it means for your bill.

49 terms defined →

Practical write-ups on cutting your AI bill - prompt caching, the batch API, self-host math.

5 guides →

Coming: the Obolith Index - tracking what it costs to run AI.