Model Status

Each model is probed with a synthetic streaming request. Latency is total completion time; TTFT is time to the first streamed token. History is kept from live probes — no synthetic data.

Loading…

auto-refreshes every 30s

Incidents

Loading…

Outcomes

Verified results — measured, not marketing

…
Tokens Served
…
Real Bugs Fixed (pytest-proven)
…
H100 GPUs Serving
…
Substantial Reasoning Outputs
…
Model Uptime (probe window)
Throughput history
tokens/min sampled every 60s from live vLLM counters — deltas of cumulative counters, no synthetic data. Chart begins when sampling started.