Skip to content

sjramblings/token-price-index

Repository files navigation

token-price-index

Machine-readable, time-versioned LLM token pricing across direct providers, aggregators, and hyperscalers.

Live URL

Target: https://tokenindex.sjramblings.io (not deployed yet — landing post-PR #5).

Project purpose

There is no machine-readable, time-versioned, multi-hyperscaler view of LLM token pricing today. Pricing pages, API catalogs, and third-party tables exist, but they are not consistently normalized into a public dataset that can be diffed over time.

The hyperscaler dimension is especially hard to see. The same model may be available directly, through Bedrock, through Vertex AI, through Azure OpenAI, or through an aggregator, but most pricing dashboards flatten that deployment context away.

This repo fills the gap as both dashboard and dataset. The public JSON files are the contract, and git history is the time-series database.

Data sources

Source URL Auth Scope
LiteLLM https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json None ~1500 models, MIT-licensed
OpenRouter https://openrouter.ai/api/v1/models None ~300 models, live API pricing
AWS Price List Bulk API https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrock/current/region_index.json None Bedrock per-region
Azure Retail Prices API https://prices.azure.com/api/retail/prices None Azure OpenAI per-region

LiteLLM is MIT-licensed; OpenRouter, AWS, and Azure terms apply to their data; this project's tooling is MIT.

How it works

Scheduled refreshes collect upstream pricing, normalize it into one schema, write a current snapshot plus a dated history file, and publish both through GitHub Pages for the dashboard SPA.

graph LR
  A[LiteLLM JSON] --> N[Normalize]
  B[OpenRouter API] --> N
  C[AWS Price List] --> N
  D[Azure Retail] --> N
  N --> CUR[data/current.json]
  N --> HIST[data/history/YYYY-MM-DD.json]
  CUR --> PAGES[GitHub Pages]
  HIST --> PAGES
  PAGES --> DASH[dashboard SPA]
Loading

Quick start

bun install
just refresh   # PR #2+ will wire this up

Roadmap

  • PR #1 — Scaffold (this PR): docs, tooling, directory structure (F1)
  • PR #2 — Data pipeline: LiteLLM + OpenRouter fetchers + Normalize + Verify (F2)
  • PR #3 — GitHub Actions daily refresh + commit-on-change + release (F3)
  • PR #4 — Vite + React dashboard with four pages (F4)
  • PR #5 — Hosting + custom domain tokenindex.sjramblings.io (F5)
  • PR #6 — Hyperscaler-pivot view (the differentiator) (F6)
  • PR #7 — Regional pricing (AWS Bedrock + Azure OpenAI per-region) (ISA § H)

License

MIT — see LICENSE

Contributing

Issues and PRs welcome.

About

Daily-refreshed unified pricing for LLM tokens across providers and hyperscalers (Bedrock, Azure OpenAI, Vertex AI). Public, machine-readable, zero-cost.

Topics

Resources

License

Stars

Watchers

Forks

Packages

 
 
 

Contributors

Languages