AI Visibility & Forecasting
Sample dataToken usage, per-model cost, and spend forecast for your Amazon Bedrock AI workloads.
Switch account
- Bedrock spend
- —
- selected range
- Proj. month-end
- —
- forecast
- Total tokens
- —
- in / out
- $ / 1K tokens
- —
- blended
- $ / inference
- —
- avg
- Cache hit
- —
- of input tokens
Error rate —
Throttle rate —
Cache reads — tok (reduced rate)
Cache writes — tok (premium)
Filter models
Token usage over time
Input vs output (aggregate) or total tokens per model.
Bedrock spend & forecast
Daily cost (solid), trend + weekday-seasonal forecast (dashed) with a 95% band. Always aggregate.
Inference requests over time
Invocations per bucket.
Tokens per inference
Average request size (input + output).
Cache token economics
Input-token mix: fresh vs cache reads (charged at a reduced rate) vs cache writes (upfront premium).
- Cache hit ratio
- —
- reads / input tokens
- Cache reads
- —
- tokens at reduced rate
- Cache writes
- —
- tokens at premium
- Fresh input
- —
- uncached input tokens
More cache reads → lower effective input cost; frequent writes without reuse erode the saving.
Spend by model
Per-model usage, cost, and unit economics. Cost shows "—" while it settles in the CUR.
| Model | Invocations | Input tok | Output tok | Total tok | Cost | $ / 1M tok | Cache hit | Cache rd | Cache wr |
|---|