OVERVIEW /All traffic
Search requests, keys, routes ⌘K
All providers
1h 24h 7d 30d
Live
Spend
$4,182
+12.0% vs yesterday
Requests
1.24M
+4.1% vs yesterday
p95 latency
1.84s
−80ms vs yesterday
Error rate
0.92%
+0.10pt vs yesterday
Cache hit rate
61.4%
+9.2pt vs yesterday
Guardrail actions
312
+38 vs yesterday
Spend $4,182.40
Anthropic OpenAI Google · other Hourly budget
Provider Model Team
$300$200$100$0
Nov 6 · 14:00
Anthropic$142.10
OpenAI$88.42
Google · other$31.90
Over budget+$22.42
Nov 5 18:0000:0006:0012:00now
Spend by model14 models
ModelCostΔ
claude-sonnet-4$1,620+$410
gpt-4o$1,104+$92
gemini-2.5-pro$712−$40
gpt-4o-mini$402+$18
llama-3.1-70b$344−$22
mistral-large$212±$0
Latency
p50 0.92s p95 1.84s p99 3.10s
Traffic 862 req/min
errors 0.92%
Top cost drivers route × model · 24h
Open in Requests
RouteModelRequestsTokensCacheCostΔ vs yesterday
support/triage claude-sonnet-4 402k 41.2M 12% $1,020 +$310 · +44%
copilot/complete gpt-4o 377k 28.7M 71% $860 +$92 · +12%
data/extract gemini-2.5-pro 118k 12.1M 64% $640 −$40 · −6%
support/summarize gpt-4o-mini 296k 30.3M 58% $510 +$18 · +4%
Alerts & findings 2 open
Budget at 92% · team support
Projected to exceed on Nov 27
4m
Cache break on support/triage
System prompt reordered · 84% → 12%
18m
31 unused tools loaded every turn
Prune to save ~29.9k tokens / call
1h
p95 > 3s · gemini-2.5-pro
Resolved after fallback to gpt-4o
2h