Qwen3.5-9B adapted — first final-answer token p95
A 9B local model answered in under a second at p95 on the diagnostic set — but did not pass behaviour gates.
Single reading
939.5 ms
Protocol
Test definition
Nearest-rank p95 over 60 frozen diagnostic cases after one-token warm-up; 4,096-token context; no truncations or timeouts.
Limits
Limitations
- Not controlled/uncontended; target is ≤ 3,000 ms; this model failed behavioural gates.
Log
Recorded values
| Date | Value | Evidence | Samples | Source |
|---|---|---|---|---|
| 21 September 2026 | 939.5 ms | MEASURED | 60 | docs/ledger/FINAL_LM_QWEN35_BOUNDED_ADAPTER_RESULT.json |
Source paths refer to the private JARVIS repository and runtime records. They are listed for audit and are not published.