I deployed an Opus 4.8 endpoint in an Azure Foundry resource in my Azure subscription. I don't understand the Opus 4.8 expense I am getting billed for on Azure.
As per Azure cost management page, I spent ~150 USD on June 26, ~150 USD on June 27 and ~200 USD on June 28 (total: 500 USD):
Details:
| Service tier | Meter | Cost |
|---|---|---|
| Claude Opus 4.8 | Claude Opus 4.8 - anthropic-claude-opus-4-8-plan - paygo-inf-longctx-cache-hit-tokens | $171.51 |
| Claude Opus 4.8 | Claude Opus 4.8 - anthropic-claude-opus-4-8-plan - paygo-inf-longctx-cache-write-tokens | $110.34 |
| Claude Opus 4.8 | Claude Opus 4.8 - anthropic-claude-opus-4-8-plan - paygo-inf-input-tokens | $71.43 |
| Claude Opus 4.8 | Claude Opus 4.8 - anthropic-claude-opus-4-8-plan - paygo-inf-output-tokens | $56.49 |
| Claude Opus 4.8 | Claude Opus 4.8 - anthropic-claude-opus-4-8-plan - paygo-inf-cache-write-tokens | $39.34 |
| Claude Opus 4.8 | Claude Opus 4.8 - anthropic-claude-opus-4-8-plan - paygo-inf-cache-hit-tokens | $35.52 |
| Claude Opus 4.8 | Claude Opus 4.8 - anthropic-claude-opus-4-8-plan - paygo-inf-longctx-output-tokens | $20.80 |
| Claude Opus 4.8 | Claude Opus 4.8 - anthropic-claude-opus-4-8-plan - paygo-inf-longctx-input-tokens | $0.54 |
| Total | $505.97 |
However, when I look at the mentoring page of the model, the token count is too low to account for 500 USD:
Cost computation:
| Type | Tokens | Rate | Cost |
|---|---|---|---|
| Input | 14.39M | $5.00/M | $71.95 |
| Output | 3.09M | $25.00/M | $77.25 |
| Total | 17.49M | $149.20 |
That's actually the most pessimistic price computation as I'm ignoring caching. As per https://marketplace.microsoft.com/en-us/product/saas/anthropic.anthropic-claude-opus-4-8-offer?tab=PlansAndPrice, it should be even cheaper since caching is used:
| Token Type | Price per 1M Tokens |
|---|---|
| paygo-inf-output-tokens | $25.00 |
| paygo-inf-longctx-output-tokens | $25.00 |
| paygo-inf-input-tokens | $5.00 |
| paygo-inf-longctx-input-tokens | $5.00 |
| paygo-inf-cache-write-tokens | $6.25 |
| paygo-inf-longctx-cache-write-tokens | $6.25 |
| paygo-inf-cache-write-1h-tokens | $10.00 |
| paygo-inf-longctx-cache-write-1h-tokens | $10.00 |
| paygo-inf-cache-hit-tokens | $0.50 |
| paygo-inf-longctx-cache-hit-tokens | $0.50 |
So 150 USD is an upper bound. Yet I got charged 500 USD. Why?