Qwen3.5-9B adapted — sustained decode throughput

About 18 tokens per second of sustained local generation.

Single reading

18.1 tokens/s

MEASURED recorded 21 September 2026

Protocol

Test definition

Tokens/s excluding prompt-to-first-token interval, over 60 cases (end-to-end aggregate 13.0 tokens/s).

Limits

Limitations

  • Not uncontended; envelope target ≥ 15 tokens/s.

Log

Recorded values

Date Value Evidence Samples Source
21 September 2026 18.1 tokens/s 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.