Inference markets, measured.
Continuous benchmarks of LLM serving — latency candles, throughput, and quality scored the way a trading desk measures execution. Every number reproducible from public raw data.
0
backends tracked
0
requests measured
—
median success
Scoreboard
The first suite publication is not live yet. This page is the public desk —
numbers appear here when ticks exist, not before.
| Backend | Model | tok/s | TTFT p50 | Requests | Success |
|---|---|---|---|---|---|
| No ticks in the store. | |||||
Placeholder until site-gen publishes from measured data · how we measure