Sample data. Not a customer bill.
Audit report
BuiltEstimated savings by lever, plus price trend, share and export.
Estimate only
This report is an estimate, not a bill. Realised savings and a count against your own baseline are planned.
August cost audit
20 Aug 2026 10:00 UTC
Estimated savings
$19.2Kest.
By lever
- Right-size support reply to Haiku1Right-size the model
Route support-reply
Estimated $4.2Kest. a month
- Trim repeated instructions on document extract5Prompt and context trimming
Route doc-extract
Estimated $1.8Kest. a month
- Send eval backfill to the batch API7Batch routing
Route eval-backfill
Estimated $900.00est. a month
- Cascade ticket triage from mini to GPT-4.12Cascade
Route ticket-triage
Estimated $2.7Kest. a month
- Reorder ticket triage so the stable prefix is cached3Provider prompt caching
Route ticket-triage
Estimated $2.7Kest. a month
- Exact-match cache for repeated support macros4Response cache
Route support-reply
Estimated $4.8Kest. a month
- Cap document extract output and require JSON6Output control
Route doc-extract
Estimated $2.1Kest. a month
Price trend
Sample prices, not list prices. Versions are kept and never overwritten, so a falling token price does not rewrite an older saving.
| Model | Host | Earlier input | Later input |
|---|---|---|---|
| GPT-4.1 | OpenAI | prices 2026-06-01 $2.50 | prices 2026-09-01 $2.00 |
| GPT-4.1 mini | OpenAI | prices 2026-06-01 $0.50 | prices 2026-09-01 $0.40 |
| Claude Sonnet 4.5 | Anthropic | prices 2026-06-01 $3.00 | prices 2026-09-01 $3.00 |
| Claude Haiku 4.5 | Anthropic | prices 2026-06-01 $1.00 | prices 2026-09-01 $1.00 |
| Gemini 2.5 Pro | Google Gemini | prices 2026-06-01 $1.50 | prices 2026-09-01 $1.25 |
| Gemini 2.5 Flash | Google Gemini | prices 2026-06-01 $0.30 | prices 2026-09-01 $0.15 |
| Mistral Large | Mistral | prices 2026-06-01 $3.00 | prices 2026-09-01 $2.00 |
| DeepSeek V3 | DeepSeek | prices 2026-06-01 $0.50 | prices 2026-09-01 $0.27 |
| Llama 3.3 70B | AWS Bedrock | prices 2026-06-01 $0.90 | prices 2026-09-01 $0.72 |