Hosted model first token p50
The hosted model's first token arrived in about half a second, slower than the design budget.
Single reading
481.8 ms
Protocol
Test definition
probe_live_providers: 10 real chat-completion calls to the pinned hosted model, prompt hand-off to first token.
Limits
Limitations
- n=10 (p95 equals max 558.2 ms); over the 320 ms p95 budget; network-dependent; local gitignored artifact.
Log
Recorded values
| Date | Value | Evidence | Samples | Source |
|---|---|---|---|---|
| 1 August 2026 | 481.8 ms | MEASURED | 10 | evidence/live/provider_pass.json |
Source paths refer to the private JARVIS repository and runtime records. They are listed for audit and are not published.