Admin · AI
LLM cost dashboard
Token spend by model and capability · 7-day view
Spend last 7d
$19
4.9M tokens
Today's spend
$5
691 requests
Monthly budget
$4K
Alert at 80%
Budget utilization
2.1%
Healthy
$85 projectedSpend by capability
Last 7 days · grouped by AI capability
Embedding$11 · 1,760 req
Contract Intel$5 · 88 req
Vendor Intel$1 · 131 req
Document Gen$1 · 690 req
Brief Intake$1 · 58 req
General$0 · 384 req
Spend by model
4 models in production rotation
- GPT-5$11 · 58.2%
- Claude Opus 4.7$5 · 27.6%
- Claude Sonnet 4.6$2 · 8.3%
- Claude Haiku 4.5$1 · 5.8%
Daily trend
Total cost per day across the last 7 days
Top capabilities by request count
Volume vs. unit economics across capabilities
| Capability | Requests | Total tokens | Cost | Avg cost / request |
|---|---|---|---|---|
| Embedding | 1,760 | 2.3M | $11 | $0.0064 |
| Document Gen | 690 | 1.5M | $1 | $0.0012 |
| General | 384 | 612K | $0 | $0.0008 |
| Vendor Intel | 131 | 201K | $1 | $0.0079 |
| Contract Intel | 88 | 179K | $5 | $0.0611 |
| Brief Intake | 58 | 107K | $1 | $0.0101 |
Token efficiency
Document gen at 5.8% of total cost generates 34.5% of requests — Haiku optimization is paying off.
Across the rotation, 3,111 requests cost $19 over 7 days — averaging $0.0063 per request. Routing high-volume document generation to Haiku keeps unit cost flat as request volume scales; Opus is reserved for contract intel where reasoning depth justifies the premium.