# Card 12 Pending Commands (Deferred to Post-Card-14 Qualification Window)

Per user decision (2026-10-31), functional harness runs, prompt lookup benchmarks, speed benchmarks,
and matched-timing comparisons on main are deferred to the single consolidated qualification window
after Card 14 completes. They must NOT be run during Card 12.

Status: PENDING
Scheduled by: Orchestrator post-Card-14

## Deferred Functional Harness Commands

1. Qwen 27B functional test suite:
   python3 tests/functional/run.py --model qwen27b --suite standard

2. Qwen 27B long-context continuation:
   python3 tests/functional/run.py --model qwen27b --suite long-context

3. Qwen 38B Flash-Next functional test suite:
   python3 tests/functional/run.py --model qwen38_flash_next --suite standard

4. Speculative prompt lookup verification:
   python3 tests/functional/prompt_lookup.py --model qwen38_flash_next

5. DeepSeek functional verification (when weights scheduled):
   python3 tests/functional/run.py --model deepseek_v4_flash --suite standard

6. Standard HTTP speed benchmark:
   python3 tools/bench/speed_bench.py --url http://127.0.0.1:8080/v1 --model qwen27b

7. Matched main-control timings (per request and phase at 5% and 10 ms):
   python3 tests/functional/run.py --model qwen27b --suite standard --compare-git-rev main
