Search for pages, actions, and quick links.
Requests, tokens, latency and spend for the assistant.
Requests (14d)
28k
Tokens (14d)
11.2kk
Spend (14d)
$337
p95 latency
3.0s
Daily totals, last 14 days
Request count by duration, with percentile markers
Total spend against cost per 1k tokens
| Model | Requests | Spend | Cost / 1k |
|---|---|---|---|
| gpt-4o-mini | 18,400 | $62.40 | $0.0006 |
| gpt-4o | 6,120 | $214.80 | $0.0075 |
| claude-sonnet | 3,980 | $141.20 | $0.0042 |
| gpt-4o (vision) | 610 | $48.60 | $0.0189 |
Figures are fixtures shaped to look like production traffic — latency in particular is right-skewed, because that is what LLM latency does, and a symmetric distribution would put p95 next to p50 and argue against its own point. Percentiles are computed from the buckets rather than written down, so the markers cannot drift from the bars.