Dragan Spiridonov db2db9dab8 feat(free-tier): default local test-gen model qwen3:30b-a3b -> qwen3-coder:30b ...

2026-06-29 oracle bench (8 fixtures, mutation>=60%) on the same Mac Ollama:
qwen3-coder:30b and qwen3:30b-a3b score IDENTICALLY (both 6/8, 75% avg
mutation kill, same fixtures pass/fail) — same 30B MoE class and RAM — but
the code-tuned model is ~14x faster wall-clock (5.2s vs 71.1s/gen) because
it does not spend ~70s "thinking" before answering.

Records both measurements in value-score.ts (incl. stock qwen2.5-coder:1.5b
= 1/8, confirming size still dominates coder-tuning at tiny scale). qwen3:8b
stays off-default (3/8 across the same fixtures).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

2026-06-29 13:16:40 +00:00

..

advisor

fix(adr-093): address devil's-advocate C1-C4 critical findings

2026-04-17 16:31:24 +00:00

escalation

fix(tests): add afterEach cleanup to 104 test files missing it (#374 P2)

2026-03-21 12:51:47 +00:00

free-tier

feat(free-tier): default local test-gen model qwen3:30b-a3b -> qwen3-coder:30b

2026-06-29 13:16:40 +00:00

security

fix(adr-093): address devil's-advocate C1-C4 critical findings

2026-04-17 16:31:24 +00:00