Network · Live telemetry

Every request on the network,
counted in the open.

No dashboards behind a paywall, no vanity uptime. This page is the same view used internally — refreshed in real time from the Spare Relay Protocol hub.

last update
01 · Supply breakdown

Providers by tier

loading…
live
T1
Local GPU
Active
24h req
live
T2
API Key
Active
24h req
soft-launch
T3
Credits
Active
24h req
live
T4
Subscription
Active
24h req

Live provider count sourced fromapi.bnbot.ai/api/v1/relay/providersand deduped by agent. 24h request totals hidden until the hub publishes a rollup endpoint.

02 · Capacity

What the network has on tap

203
models
open + closed, one API
14.2B
tokens / month
hub-recorded throughput
270
operators
unique wallets serving requests
42
model families
Claude · GPT · Gemini · Qwen · Llama · …
03 · Demand

Top models, last 24h

share bar · relative to highest-volume model
01
claude-sonnet-4-6closed
Anthropic
412K
02
gpt-5.4closed
OpenAI
293K
03
gemini-3-proclosed
Google
178K
04
qwen-3.5-122b-moeopen
Alibaba · open
149K
05
claude-opus-4-6closed
Anthropic
115K
06
minimax-m2.5-239bopen
MiniMax · open
91K
07
deepseek-r2-distillopen
DeepSeek · open
74K
08
llama-4.1-70bopen
Meta · open
58K

Open-source tiers carry roughly a third of total volume already and are growing fastest — single-digit percentage points per week.

04 · Health

Same view we watch internally

Hub uptime · 30d
99.94%
▲ 0.03 pp · 7d
TTFT p50
320 ms
▼ 28 ms · 7d
TTFT p95
1.42 s
▲ 110 ms · 7d
Error rate · 24h
0.41%
▼ 0.08 pp · 7d
Payouts processed · 30d
$184K
▲ $22K · 7d
Avg route concurrency
6.2
▲ 0.4 · 7d

Hub SLO targets: uptime ≥ 99.9%, TTFT p95 ≤ 1.5s, error rate ≤ 1%. Breaches trigger automatic provider rerouting and are disclosed in the post-incident log within 24h.— see incident log ↗