FinOps dashboard
How much your Sentinel fleet spends in tokens, and where they go
The FinOps dashboard shows how your fleet consumes tokens — where they go and how efficiently they're used — so you can spot waste and tune your AI spend. Find it in the SUPERWISE® app under Observability → Dashboards → FinOps.
Use the sentinel filter to focus on specific gateways, the time range (24H / 7D / 30D) to set the window, and the Live toggle to auto-refresh. Every metric compares against the previous period, and you can export the view as a PDF, PNG, or CSV.
Like every Sentinel dashboard, this is computed from operational metadata only: token counts and model names, never your prompts or responses. See Data privacy.
The dashboard has three sections.
Consumption
Total token usage for the period and the efficiency ratios behind it:
- Total consumption — all tokens consumed (prompt, cache read/write, and completion) across every provider and sentinel.
- Cache hit rate — the share of input tokens served from cache. Higher is cheaper.
- Reasoning ratio — reasoning tokens as a share of completion tokens, so you can see how much "thinking" your traffic pays for.
- Output-to-input ratio — completion tokens relative to input tokens.
- Consumption over time — token usage across the window.
Where do the tokens go?
The same consumption, broken down so you can find the heavy hitters:
- By provider — token usage per provider, including the cache and reasoning breakdown.
- Top models by tokens — which model versions consume the most.
- By sentinel — consumption across your gateways.
Token saving
How much token optimization is giving back, if any of your Sentinels have it on:
- Tokens saved — the total tokens Headroom and Caveman kept out of your traffic over the period, with the trend versus the previous period.
- Token saving over time — savings across the window, so you can see the effect build up.
- Compression strategy detail — a breakdown of savings by method (input compression with Headroom versus output shortening with Caveman), so you can see which is doing the work.
Headroom savings are measured (tokens before and after compression); Caveman savings are estimated from the response, since it shortens generation rather than rewriting a prompt. See Token optimization.
Tokens not CostThe FinOps dashboard measures token consumption, not billed currency. Pair it with your provider's own pricing to translate usage into cost.
Get Sentinel: Sentinel on superwise.ai.
See also
- Token optimization: compress prompts to cut token usage.
- Operational dashboard: fleet health and latency.
- Live activity: each interaction as it happens.
Updated 19 days ago