Every model you pay for, in one table

What each model really costs

Import your OpenAI and Anthropic usage exports, add DeepSeek and the Model Studio lines in your Alibaba Cloud bill, and see spend, tokens and cost per million tokens for each model, beside the rest of your cloud bill.

Free and open source for macOS and Windows. No account, no sync, no telemetry.

How it works

From a console export to cost per token

No API keys for OpenAI or Anthropic: the export their consoles give you is enough.

  1. Download the export

    OpenAI: Usage → Export. Anthropic: Usage or Cost → Export. DeepSeek: Usage → Download.

  2. Import it

    Add the account with a name and no credentials, then click Import on its row. Importing a month again replaces it.

  3. Read the Models page

    Each model's cost and change, its input, output and cache tokens, and what it costs per million tokens.

Sources

The model bills it reads

Token-metered rows from any source show up on the Models page; the model is the one the bill names.

ProviderHowWhat it adds
OpenAIUsage exportCost or token usage, per project and model
Anthropic (Claude)Usage or cost exportCost or token usage, per workspace and model
DeepSeekBalance API, usage downloadPrepaid balance; per-day, per-model spend
Alibaba Cloud Model StudioBill details exportModel Studio (百炼) costs per model
Volcengine ArkBill details exportArk (火山方舟) costs per endpoint and token type

The Models page

Token economics, per model

The stat row shows AI spend, total tokens, the blended cost per million tokens and the cache share. Below it, each model's cost and change, its input, output and cache tokens and its share of model spend, and a chart of tokens per day.

Models Try it in the demo

  • Flags a model whose cache reads are under 5% of its input side: input price paid for what could be cache price.
  • Flags one model taking more than 70% of model spend.
  • Tells a rise in unit price apart from a rise in usage.
  • Points out a model billed without token metering, whose unit cost cannot be computed.

Keep watch

Know when model spend jumps

The cost growth anomaly rule flags a source and service whose daily spend runs above its trailing 7-day baseline for two days running. The balance floor rule warns before a prepaid balance like DeepSeek's runs out. CloudBridge posts both from the menu bar.

Cloud cost alerts

  • Model spend sits in the same ledger as AWS, Alibaba Cloud and Cloudflare, so one Overview covers it all.
  • Original currencies are kept beside the converted amount; missing exchange rates are reported, not assumed.
  • The ledger is a DuckDB file on your disk, with read-only SQL on the Query page.

FAQ

Questions people ask

Do I need to give CloudBridge my OpenAI or Anthropic API key?
No. OpenAI and Anthropic are import-only: add the account with a name and no credentials, then import the cost or usage export downloaded from the provider's console.
What if my export has tokens but no cost?
Token-only exports stay usage records. CloudBridge counts their tokens on the Models page but does not turn them into an estimated bill.
Which model providers does it read?
OpenAI and Anthropic exports, DeepSeek's balance API and usage download, Alibaba Cloud Model Studio through the Alibaba Cloud bill, and Volcengine Ark through the Volcengine bill export.
Can it warn me when model spend jumps?
Yes. The cost growth anomaly rule flags daily spend above its trailing 7-day baseline for consecutive days, and the balance floor rule warns when a prepaid balance such as DeepSeek's runs low.

See your model bill per model

Download CloudBridge, import last month's exports, and open the Models page. Or look around the demo first.